Stoichiometric Modeling of BLSS Mass Flows: From Closed-Loop Systems to Clinical Applications

Abstract

Stoichiometric modeling is pivotal for designing Bioregenerative Life Support Systems (BLSS) that enable long-duration space missions by closing mass loops and providing essential resources. This article explores the foundational principles of element cycling in artificial ecosystems, detailing methodologies like Flux Balance Analysis for predicting intracellular fluxes. It addresses key challenges in model optimization and thermodynamic feasibility, and reviews advanced validation techniques to ensure predictive reliability. By synthesizing insights from space-life-support research and constraint-based metabolic modeling, we highlight the cross-disciplinary applications of these frameworks in biomedical research, including drug discovery and understanding metabolic diseases.

Principles of Mass Flow and Element Cycling in Closed Ecosystems

The Role of BLSS in Long-Duration Space Missions and Mass Closure

Bioregenerative Life Support Systems (BLSS) are advanced artificial ecosystems considered vital for future long-duration and remote space missions. These systems are designed to recycle human metabolic wastes into nutrients, carbon dioxide, and water for plants and other edible organisms, which in turn provide food, fresh water, and oxygen for astronauts. The central concept involves creating a materially closed loop to significantly reduce mission mass and volume by cutting down or even eliminating disposable waste and reliance on resupply missions from Earth [1]. For autonomous long-duration space missions without resupply possibility, a BLSS that generates all essential resources with minimal material loss is fundamental for mission sustainability [1] [2].

The core principle of a BLSS mimics natural ecological networks, comprising three main types of biological compartments: producers (e.g., plants, microalgae), consumers (i.e., crew), and degraders/recyclers (e.g., bacteria) [3]. The ultimate goal is to achieve a high degree of mass closure, where the majority of resources are regenerated within the system. Recent achievements, such as the Chinese "Lunar Palace 365" mission, have demonstrated an overall system closure degree of 98.2% over a 370-day experiment, providing strong evidence for the feasibility of this technology for future lunar bases [4].

Stoichiometric Modeling of BLSS Mass Flows

Fundamental Principles

Stoichiometric modeling provides the foundational framework for describing and quantifying the mass flows of elements within a BLSS. It involves establishing a compact set of chemical equations with fixed coefficients to describe the cycling of key elements—primarily Carbon (C), Hydrogen (H), Oxygen (O), and Nitrogen (N)—through all interconnected compartments of the system [1]. This approach allows researchers to simulate the flow of all relevant compounds and balance the dimensions of different compartments to maximize closure at steady state.

The stoichiometric relations govern the material flows in the ecosystem model, enabling the prediction of system dynamics, long-term reliability, and the impact of design changes or perturbations [1]. In a successfully balanced system, most compounds exhibit minimal or zero loss between process iterations. For instance, recent modeling efforts have demonstrated that 12 out of 14 compounds can achieve zero loss, with only oxygen and CO2 displaying minor, manageable losses [1].

MELiSSA as a Modeling Framework

The Micro-Ecological Life Support System Alternative (MELiSSA) project, developed by the European Space Agency with international partners, serves as a leading reference framework for BLSS stoichiometric modeling. The MELiSSA loop is structured as an artificial ecosystem consisting of five interconnected compartments inhabited by different organisms, each with specific metabolic functions [1]:

• C1: Thermophilic anaerobic compartment for initial waste breakdown
• C2: Photoheterotrophic compartment
• C3: Nitrifying compartment
• C4a & C4b: Photoautotrophic compartments (microalgae and higher plants)
• C5: Crew compartment (human inhabitants)

This compartmentalized approach enables specialized processing of waste streams and efficient regeneration of resources through controlled biochemical pathways, providing an ideal structure for stoichiometric analysis [1].

BLSS Compartment Protocols and Methodologies

Higher Plant Compartment (C4b) Protocol

The higher plant compartment serves as the primary producer of food, oxygen, and water transpiration while consuming CO2 and nutrients. The experimental protocol varies significantly based on mission duration and objectives.

• Application: Food production, oxygen generation, CO2 removal, water purification, psychological support [3]
• Species Selection Criteria:
  • Short-duration missions (<6 months): Fast-growing species with high nutritive value and minimal volume requirements (leafy greens, microgreens, dwarf cultivars) [3]
  • Long-duration missions (>6 months): Staple crops providing carbohydrates, proteins, and fats (wheat, potato, rice, soy), plus vegetables and fruits with longer growth cycles (~100 days) [3]

Experimental Workflow:

• Cultivation System Setup: Install controlled environment agriculture systems with precise monitoring of temperature, humidity, light intensity, photoperiod, and CO2 concentration [3].
• Nutrient Delivery: Implement hydroponic, aeroponic, or soil-like substrate systems using recycled nutrients from waste processing compartments [4].
• Growth Monitoring: Daily tracking of plant health, development stage, and environmental parameters.
• Harvest and Processing: Scheduled harvesting of edible biomass with waste biomass redirected to recycling compartments.
• Yield Analysis: Quantification of edible biomass production, nutritional content, and resource consumption rates.

Table 1: Plant Species Selection for Different Mission Scenarios

Mission Type	Example Species	Growth Cycle	Primary Output	Resource Contribution
Short-duration	Lettuce, Kale, Microgreens	20-30 days	Nutritional supplementation, antioxidants	Limited resource recycling
Long-duration	Wheat, Potato, Rice, Soy	80-120 days	Caloric and protein provision	Significant O2 production & CO2 consumption
Supplemental	Tomato, Peppers, Beans, Berries	~100 days	Dietary variety, phytonutrients	Moderate resource recycling

Microbial Waste Processing Compartments (C1-C3) Protocol

The microbial compartments are responsible for the systematic breakdown of human waste and conversion into usable nutrients for plant compartments.

• Application: Waste mineralization, nutrient recovery, CO2 production [1]
• Organisms: Anaerobic bacteria (C1), photoheterotrophic bacteria (C2), nitrifying bacteria (C3) [1]

Experimental Workflow:

• Bioreactor Inoculation: Establish pure or controlled mixed cultures of specific bacterial strains in appropriate bioreactors [1].
• Waste Feed Introduction: Introduce standardized human waste simulants or actual metabolic wastes at controlled feed rates.
• Process Monitoring: Continuously monitor temperature, pH, dissolved oxygen, and metabolic intermediates (e.g., volatile fatty acids).
• Output Characterization: Analyze effluent for nutrient content (nitrates, phosphates), elemental composition, and potential contaminants.
• System Integration: Transfer processed waste streams to appropriate downstream compartments (C3 for nitrification, C4 for plant nutrition).

Gas Balance Management Protocol

Maintaining O2 and CO2 concentrations within appropriate ranges is a critical indicator of BLSS stability and requires active management [4].

• Application: Atmospheric homeostasis, crew health protection [4]
• Target Parameters: CO2 concentration 246-4131 ppm, O2 concentration at breathable levels [4]

Experimental Workflow:

• Continuous Monitoring: Implement real-time gas sensors for O2, CO2, and trace contaminants throughout the system.
• Metabolic Load Prediction: Calculate expected O2 consumption and CO2 production based on crew activity and plant respiration cycles.
• Intervention Implementation: Deploy strategic countermeasures during crew shift changes or system perturbations:
  • Strategy A: Temporarily store excess CO2 in storage materials until photosynthetic activity increases [4]
  • Strategy B: Adjust the illuminated area for plants or manipulate light intensity to regulate photosynthetic rates [4]
• Performance Validation: Verify system recovery to equilibrium gas concentrations following interventions.

Quantitative Analysis of BLSS Performance

Table 2: Mass Closure Performance in Recent BLSS Experiments

System/Experiment	Duration	Crew Size	O2 Closure (%)	Water Closure (%)	Food Closure (%)	Overall Closure (%)
Lunar Palace 365 [4]	370 days	4 (rotating)	100%	100%	High (partial resupply)	98.2%
MELiSSA Model [1]	Steady-state simulation	6	~100% (minor losses)	Not specified	100%	High (12/14 compounds zero loss)
Early CELSS [2]	91 days	4	Significant contribution	Not specified	Partial supplementation	Not specified

Table 3: Stoichiometric Element Tracking in BLSS Modeling

Element	Input Sources	Output Sinks	Recycling Pathways	Measurement Techniques
Carbon (C)	Crew respiration (CO2), waste	Plant biomass, microbial biomass,	Photosynthesis, waste degradation	CO2 sensors, biomass composition analysis
Hydrogen (H)	Water, organic compounds	Water vapor, biomass,	Transpiration, condensation,	Mass balance, humidity sensors
Oxygen (O)	CO2, water, plant production	Crew consumption, oxidation processes	Photosynthesis, respiration	O2 sensors, gas chromatography
Nitrogen (N)	Crew waste (urea), food	Plant proteins, microbial biomass	Nitrification, assimilation	Elemental analysis, ion chromatography

Research Reagent Solutions and Essential Materials

Table 4: Key Research Reagents and Materials for BLSS Experimentation

Reagent/Material	Function/Application	Specification Requirements
Trimethylolpropane (TMP) [5]	Bio-lubricant synthesis for machinery maintenance	High purity, esterification grade
Hydroponic nutrient solutions [3]	Plant mineral nutrition	Balanced macro/micronutrients, pH buffered
Bacterial culture media [1]	Waste processor inoculation	Sterile, defined composition for target microbes
Gas standard mixtures [4]	Sensor calibration	Certified O2, CO2, trace contaminants in balance gas
Water quality test kits [6]	Monitoring recycled water safety	Tests for microbial contamination, organics, ions

BLSS System Visualization and Workflows

BLSS Compartment Mass Flow Diagram

BLSS Mass Flow Diagram

Stoichiometric Modeling Workflow

Stoichiometric Modeling Workflow

Bioregenerative Life Support Systems with sophisticated stoichiometric modeling represent the pinnacle of life support technology for long-duration space missions. The integration of biological components with engineering controls enables unprecedented levels of mass closure, reducing reliance on Earth resupply. Current research demonstrates that over 98% closure is achievable in ground-based demonstrations, with mathematical models supporting the feasibility of fully autonomous systems [4] [1].

Future development should focus on enhancing system resilience to perturbations, particularly during crew shift changes, improving the oxidation stability of biological components, and validating system performance under actual space conditions [2] [4]. As space agencies worldwide prepare for sustained lunar presence and eventual Mars exploration, BLSS technology with robust stoichiometric modeling will be fundamental to mission success and crew survival in the challenging environment of deep space.

The Micro-Ecological Life Support System Alternative (MELiSSA) is an artificial ecosystem conceived as a tool for understanding the behavior of closed-loop biological systems and developing technology for future biological life support systems (BLSS) in long-term space missions [7]. The primary objective of MELiSSA is the recovery of oxygen and edible biomass from waste materials, including faeces and urea [7]. Due to the intrinsic instability of such complex biological systems and the stringent safety requirements of manned space missions, a sophisticated hierarchical control strategy has been developed to pilot the system and optimize its recycling performance [7]. The framework is structured as an assembly of unit processes, or compartments, designed to simplify the behavior of the artificial ecosystem and enable a deterministic engineering approach [8]. This organization into specific compartments with assigned functions allows for detailed stoichiometric modeling of mass flows, which is fundamental to the research and development of BLSS.

The Five-Compartment Structure

The MELiSSA loop is engineered as a sequential process where the output of one compartment serves as the input for the next, ultimately supporting human life. The specific functions of each compartment are detailed in Table 1.

Table 1: The Five Compartments of the MELiSSA Loop

Compartment	Key Microorganisms / Components	Primary Function	Key Inputs	Key Outputs
CI (Liquefying Compartment)	Thermophilic anoxygenic bacteria [8]	Organic waste degradation & solubilisation [8]	Organic wastes (e.g., non-edible plant parts, paper) [8]	CO₂, volatile fatty acids, ammonia [8]
CII (Photoheterotrophic Compartment)	Photoheterotrophic bacteria [8]	Removal of organic carbon compounds [8]	Volatile fatty acids, ammonia from CI [8]	Inorganic carbon source [8]
CIII (Nitrifying Compartment)	Nitrosomonas europaea, Nitrobacter winogradskyi (in a biofilm) [8]	Conversion of ammonia into nitrates [8]	Ammonia from preceding compartments [8]	Nitrates (suitable nitrogen source for plants) [8]
CIVa (Photoautotrophic Compartment - Bacteria)	Arthrospira platensis (cyanobacteria) [8]	Food and oxygen production [8]	CO₂ from CI and crew, nutrients [8]	Edible biomass, oxygen, water [8]
CIVb (Photoautotrophic Compartment - Higher Plants)	Higher plants (e.g., Lactuca sativa; 32 crops considered) [8] [9]	Food, oxygen, and water production [8]	CO₂ from CI and crew, nitrates from CIII [8]	Edible biomass, oxygen, water [8]
CV (Crew Compartment)	Human crew	Consumption of resources and production of waste	O₂, food, water from CIVa and CIVb [8]	CO₂, organic waste, urea [7]

The logical flow and mass exchange between these compartments and the crew can be visualized as a circular ecosystem.

Stoichiometric Modeling and Control Strategies

The driving element of MELiSSA is the efficient recovery of mass and energy. A hierarchical control strategy is employed to ensure system stability and performance [7]. This strategy operates at two primary levels:

• Local Control: Each MELiSSA compartment has its own local control system [7].
• Global Control: An upper-level control system, taking into account the states of all compartments and a desired global functioning point, determines the optimal setpoints for each local controller [7].

This approach is fundamentally based on first principles models of each compartment, which incorporate physico-chemical equations, stoichiometries, and kinetic rates [7]. These models are used both for developing a global system simulator and for implementing a non-linear predictive model-based control strategy [7]. For higher plant chambers (Compartment IVb), modeling is particularly complex. A multilevel mechanistic modeling approach has been developed to integrate phenomena across different scales, from the canopy level down to the metabolic network [9]. This approach can include Flux Balance Analysis (FBA) to predict the distribution of metabolic fluxes, providing a deeper understanding of the plant's internal stoichiometry and its response to environmental conditions [9]. The integration of these detailed models allows for the development of advanced Model Predictive Control (MPC) architectures that can manage the chamber environment to optimize plant growth and system-level mass flows [9].

Experimental Protocols for System Validation

Protocol: Global Control Strategy Simulation and Validation

This protocol outlines the procedure for using first-principles models to simulate the global MELiSSA ecosystem and validate its control strategy [7].

1. Objective: To simulate the dynamic behavior of the interconnected MELiSSA loop and validate the hierarchical control strategy's ability to maintain system stability and performance at a defined global functioning point.

2. Research Reagent Solutions and Essential Materials: Table 2: Key Materials for MELiSSA Research

Item / Organism	Function in the Ecosystem	Research Context
Thermophilic anoxygenic bacteria	Degrades solid organic waste into soluble compounds in CI [8].	Used in bioreactor studies for waste liquefaction efficiency.
Photoheterotrophic bacteria	Removes organic carbon compounds from CI effluent in CII [8].	Key for preventing feedback inhibition in CI.
Nitrifying bacteria consortium (Nitrosomonas europaea, Nitrobacter winogradskyi)	Converts toxic ammonia into nitrate, the preferred nitrogen source for plants, in CIII [8].	Essential for nitrogen cycle closure.
Arthrospira platensis (Cyanobacteria)	Produces oxygen, edible biomass, and water through photosynthesis in CIVa [8].	Studied for its high growth rate and nutritional value.
Lactuca sativa (Lettuce) and other higher plants	Produces a varied diet, oxygen, and water, and contributes to well-being in CIVb [9].	Model organism for higher plant chamber research.
First-Principles Compartment Models	Mathematical models containing physico-chemical equations, stoichiometries, and kinetic rates [7].	Core component of the global simulator and predictive controller.

3. Methodology: 1. Model Integration: Develop or obtain the validated first-principles models for each of the five MELiSSA compartments (CI, CII, CIII, CIVa, CIVb) and the crew (CV). These models should encapsulate the core stoichiometries and kinetics of the biological processes [7]. 2. Simulator Configuration: Integrate the individual compartment models into a global simulator. The outputs of one compartment (e.g., CO₂ from CI and CV) must be correctly linked as inputs to the downstream compartments (e.g., CIVa and CIVb) [7] [8]. 3. Control System Implementation: - Implement the local controllers for each compartment, which regulate internal parameters based on local setpoints. - Implement the upper-level global controller, which uses the global simulator in a predictive manner. This controller monitors the state of all compartments and calculates new setpoints for the local controllers to drive the system towards an optimal, safe operating point [7]. 4. Simulation Execution: Run the coupled simulation and control system over a defined mission period. Introduce realistic perturbations, such as a variation in crew waste output or a change in light intensity for the photosynthetic compartments. 5. Data Collection and Analysis: Monitor key performance indicators (KPIs) including: - Oxygen and carbon dioxide levels. - Production rates of edible biomass. - Stability of each compartment's key process variables. - Overall mass flow closure.

4. Anticipated Outcomes: The simulation will demonstrate whether the hierarchical control strategy can successfully reject disturbances and maintain the entire MELiSSA loop at the desired recycling performance, thereby validating the control approach before implementation in a physical pilot plant.

Protocol: Multilevel Modeling and Control of a Higher Plant Chamber

This protocol details the methodology for developing and validating a multilevel model for higher plant growth (CIVb) and integrating it into a model-based predictive controller [9].

1. Objective: To create and validate a mechanistic multilevel model of Lactuca sativa (lettuce) growth and use it to design a predictive controller for optimizing environmental conditions in the plant chamber.

2. Methodology: 1. Model Development (Multilevel Approach): - Level 1 (Canopy/Chamber Scale): Develop sub-models for irradiance distribution within the canopy, energy balance (to determine leaf temperature), and gas exchange (CO₂, O₂, H₂O) between the plant and the chamber atmosphere [9]. - Level 2 (Biochemical Level): Implement enzyme-kinetic based models for fundamental processes like photosynthesis (e.g., the Farquhar model) and respiration [9]. - Level 3 (Metabolic Network Level): Reconstruct a genome-scale metabolic network for Lactuca sativa. Use Flux Balance Analysis (FBA) to predict intracellular flux distributions and growth rates under the constraints provided by Level 1 and 2 models [9]. 2. Model Validation: Grow Lactuca sativa in a controlled environment chamber. Collect experimental data on gas exchange rates (CO₂ uptake, O₂ production, water transpiration), biomass accumulation, and environmental conditions (light, temperature, humidity). Compare these measurements against the predictions of the multilevel model to validate its accuracy [9]. 3. Controller Design: Embed the validated multilevel model into a Model Predictive Control (MPC) framework. The MPC algorithm will use the model to predict future plant growth and gas exchange based on current states. It will then compute optimal adjustments to the chamber's control variables (e.g., light intensity, CO₂ concentration, irrigation) to maximize a predefined objective, such as biomass production rate or oxygen regeneration [9]. 4. Experimental Control Validation: Implement the MPC system on the actual plant growth chamber and run a controlled experiment. Compare the system's performance (e.g., growth rate, resource use efficiency) against traditional control strategies.

3. Anticipated Outcomes: This protocol enables a deeper, mechanistic understanding of plant growth in controlled environments. The resulting MPC strategy is anticipated to outperform traditional controllers, leading to more precise and efficient management of the photoautotrophic compartment (CIVb), which is critical for the overall success of a BLSS [9].

Bioregenerative Life Support Systems (BLSS) are artificial ecosystems critical for long-duration space missions, as they recycle human waste into oxygen, water, and food, thereby creating a materially closed loop and reducing mission mass and volume [1]. Stoichiometric modeling forms the foundational framework for understanding and predicting the mass flows of key elements—Carbon (C), Hydrogen (H), Oxygen (O), and Nitrogen (N)—through these systems. The accurate balancing of these elements is paramount for achieving a high degree of closure and ensuring the continuous, sustainable provision of vital resources for the crew without external resupply [1]. This document outlines the core stoichiometric equations and associated protocols for modeling the mass flows of C, H, O, and N within a BLSS, based on the established MELiSSA (Micro-Ecological Life Support System Alternative) concept.

The MELiSSA loop, developed by the European Space Agency, is a benchmark BLSS architecture composed of five distinct, interconnected compartments, each with a specific metabolic function [1]. The system's operation relies on the sequential processing of waste by various organisms to ultimately sustain human life. A conceptual model of the mass flows between these compartments is provided in the diagram below.

Core Stoichiometric Equations by Compartment

The following section details the fundamental stoichiometric equations for the cycling of C, H, O, and N in each compartment. These equations are based on a simplified, balanced model designed for a crew of six and assume steady-state operation [1].

Table 1: Core Stoichiometric Equations for the MELiSSA Loop [1]

Compartment	Primary Function	Core Stoichiometric Equation
C1	Thermophilic Anaerobic Digestion	Organic Waste (CxHyOz) + H₂O → Volatile Fatty Acids (VFAs) + CO₂ + CH₄ + NH₄⁺ + ...
C2	Photoheterotrophic Oxidation	VFAs (e.g., C₂H₄O₂) + O₂ + NH₄⁺ → Bacterial Biomass (C₅H₇O₂N) + CO₂ + H₂O
C3	Nitrification	NH₄⁺ + 1.5 O₂ → NO₂⁻ + 2H⁺ + H₂ONO₂⁻ + 0.5 O₂ → NO₃⁻
C4a	Microalgae (e.g., Limnospira)	CO₂ + NO₃⁻ + H₂O + Light → Algal Biomass (C₆H₁₀O₅N) + O₂
C4b	Higher Plants (e.g., Wheat)	CO₂ + NO₃⁻ + H₂O + Light → Plant Biomass (C₆H₁₂O₆) + O₂
C5	Human Crew	Food (C₆H₁₂O₆, C₆H₁₀O₅N, etc.) + O₂ → CO₂ + H₂O + Urea (CH₄N₂O) + Feces

The logical sequence of these chemical transformations, which close the elemental loops, is visualized below.

Quantitative Data for a 6-Person Crew

For practical system design, the stoichiometric model must be quantified. The table below summarizes the key mass flows for a system supporting a crew of six, based on a balanced steady-state model [1].

Table 2: Key Mass Flow Rates for a 6-Person Crew in a Closed BLSS [1]

Compound / Element	Mass Flow Rate (g/day)	Source Compartment	Sink Compartment	Notes
O₂ (Oxygen)	~2000	C4a, C4b	C5	Primary product of photosynthesis; consumed by crew.
CO₂ (Carbon Dioxide)	~2500	C5, C1, C2	C4a, C4b	Primary product of respiration and breakdown; consumed by plants/microalgae.
H₂O (Water)	Variable	All	All	Recycled and purified throughout the loop.
Edible Biomass	~1500 (dry weight)	C4a, C4b	C5	Provides 100% of crew's nutritional needs in a fully closed system.
Nitrate (NO₃⁻)	To be balanced	C3	C4a, C4b	Key nitrogen source for photoautotrophs.
Ammonium (NH₄⁺)	To be balanced	C1, C5	C2, C3	Key intermediate in nitrogen cycle.

Experimental Protocols for Stoichiometric Model Calibration

Protocol: Determination of Crew Metabolic Stoichiometry

Objective: To empirically determine the consumption and production rates of C, H, O, and N for a human in a controlled environment, providing the foundational input (C5) for the BLSS model.

Materials:

• Calorimetry chamber (whole-room or mask-based)
• Gas analyzers (for O₂ and CO₂)
• Automated urine and feces collection system
• Food and water with precisely known elemental composition
• Microbalance for accurate mass measurements

Methodology:

• Subject Preparation: Subjects reside in a sealed calorimetry chamber for a minimum of 72 hours.
• Controlled Diet: Provide subjects with a diet of known mass and stoichiometric composition (CxHyOzN). Record all food and water intake mass.
• Gas Exchange Monitoring: Continuously monitor the concentration and flow rate of inlet and outlet air to calculate real-time O₂ consumption and CO₂ production rates.
• Waste Collection: Collect all urine and feces excreted during the study period. Record mass.
• Sample Analysis:
  • Elemental Analysis: Use CHNS elemental analysis on aliquots of homogenized food, feces, and urine to determine C, H, N content.
  • Calorimetry: Use bomb calorimetry to cross-validate energy content, which correlates with carbon oxidation.
• Data Calculation:
  • Calculate daily intake of each element (from food, water, O₂).
  • Calculate daily output of each element (in CO₂, feces, urine, H₂O).
  • Formulate an empirical chemical equation representing human metabolism, balancing input and output masses for C, H, O, and N.

Protocol: Characterization of Photobioreactor (C4a) Stoichiometry

Objective: To establish the growth stoichiometry of Limnospira indica microalgae under defined light and nutrient conditions, quantifying its O₂ production and nutrient uptake.

Materials:

• Photobioreactor (PBR) with controlled temperature, pH, and lighting
• Sterile culture medium with known NO₃⁻ concentration
• CO₂ supply system with mass flow controller
• In-situ optical density (OD) probe or cell counter
• Gas analyzer for outlet O₂ and CO₂
• Filtration setup for biomass harvesting

Methodology:

• Inoculation: Aseptically inoculate the PBR with an axenic culture of L. indica.
• Continuous Cultivation: Operate the PBR in continuous or semi-continuous mode to achieve steady-state growth.
• Process Monitoring: Continuously log light intensity, temperature, pH, and OD.
• Gas Analysis: Precisely measure the inflow rate of CO₂ and the outflow rates of O₂ and residual CO₂.
• Medium & Biomass Sampling:
  • Take periodic samples from the culture medium. Use ion chromatography to measure NO₃⁻ depletion over time.
  • Harvest a known volume of culture, filter, and wash the biomass.
• Biomass Analysis:
  • Dry Weight: Determine the biomass dry weight.
  • Elemental Analysis: Perform CHNS analysis on the dried biomass to determine its empirical formula (e.g., ~C₆H₁₀O₅N).
• Stoichiometry Calculation:
  • Using the data on CO₂ consumed, O₂ produced, NO₃⁻ consumed, and biomass produced, derive the balanced stoichiometric equation for microalgae growth under the tested conditions.

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents and Materials for BLSS Stoichiometric Research

Item Name	Function / Application	Critical Specifications
CHNS Elemental Analyzer	Precisely determines the mass fractions of Carbon, Hydrogen, Nitrogen, and Sulfur in solid and liquid samples (e.g., biomass, food, waste).	High accuracy (±0.3%), ability to handle small sample masses (1-3 mg).
Gas Chromatography (GC) System	Separates and quantifies gas mixtures; essential for monitoring CH₄, CO₂, and O₂ in the headspace of anaerobic (C1) and aerobic (C2, C4) reactors.	Equipped with TCD (Thermal Conductivity Detector) and FID (Flame Ionization Detector).
Ion Chromatography (IC) System	Measures concentrations of anions (NO₃⁻, NO₂⁻, PO₄³⁻) and cations (NH₄⁺, K⁺, Ca²⁺) in liquid culture media, tracking nutrient uptake and conversion.	High sensitivity in the parts-per-million (ppm) range.
Photobioreactor (PBR)	Provides a controlled environment (light, temperature, pH, gas mixing) for cultivating microalgae (C4a) and characterizing its growth stoichiometry.	Integrated sensors for pH, dO₂, temperature; adjustable light intensity.
Synthetic Waste Stream	A chemically defined simulant of human waste used for reproducible experimentation with C1 and C2 compartments, avoiding variability of real waste.	Precise composition of carbohydrates, proteins, lipids, and minerals based on human metabolic studies.
Limnospira indica Culture	A model cyanobacterium for the C4a compartment; efficiently produces O₂ and edible biomass from CO₂ and minerals.	Axenic (sterile, contaminant-free), high growth rate strain.

Defining Empirical Formulas for Key Biomolecules and Biomass

In the context of stoichiometric modeling of Bioregenerative Life Support Systems (BLSS), the precise definition of empirical formulas for biomolecules and bulk biomass is not merely an analytical exercise but a foundational requirement for predicting mass flows and achieving system closure. A BLSS aims to recycle astronaut metabolic waste into food, oxygen, and clean water through a series of interconnected biological compartments [1]. Accurate stoichiometric models, which track the flow of elements like carbon, hydrogen, oxygen, and nitrogen (CHON), are essential for designing these systems to be sustainable for long-duration space missions without resupply [1]. The empirical formula, which expresses the simplest whole-number ratio of atoms of each element in a compound, provides the fundamental building block for these complex mass balance calculations [10]. This document outlines the theoretical principles and detailed experimental protocols for determining these critical values, enabling researchers to construct reliable models for BLSS research and development.

Fundamental Concepts: Empirical vs. Molecular Formulas

Understanding the distinction between empirical and molecular formulas is crucial for accurate stoichiometric accounting.

• Empirical Formula: Represents the simplest whole-number ratio of the atoms of each element present in a compound. It is a standardized descriptor of composition. For example, the empirical formula for glucose is CH₂O, indicating a 1:2:1 ratio of carbon, hydrogen, and oxygen atoms [10] [11].
• Molecular Formula: Indicates the actual number of atoms of each element in a single molecule of the compound. The molecular formula is always a whole-number multiple of the empirical formula. For glucose, the molecular formula is C₆H₁₂O₆, which is six times its empirical formula [10].

For stoichiometric modeling of processes like biomass combustion or microbial digestion, the empirical formula is often sufficient as it defines the fundamental elemental ratio being transformed [10].

Analytical Protocols for Compositional Analysis

Determining the empirical formula of a homogeneous biomolecule follows a well-established calculative procedure, while defining a representative formula for heterogeneous biomass requires extensive analytical characterization.

Protocol 1: Determining the Empirical Formula from Percentage Composition

This method is suitable for purified compounds [10] [11].

Workflow Overview:

Detailed Procedure:

• Obtain Percentage Composition: Acquire data on the mass percentage of each element (C, H, O, N, etc.) in the compound through elemental analysis. The sum should be 100%.
• Assume a 100g Sample: Simplify calculations by assuming a 100-gram sample. This converts percentage values directly into gram masses (e.g., 40% C becomes 40g C).
• Convert Mass to Moles: For each element, divide the mass by its atomic molar mass.
  • Example: For 40g of Carbon: Moles of C = 40 g / 12.01 g/mol ≈ 3.33 mol [10].
• Determine the Simplest Ratio: Divide all mole values by the smallest number of moles calculated in the previous step.
• Achieve Whole Numbers: If the ratios are not whole numbers (e.g., 1.33), multiply all ratios by a small integer (e.g., 3) to obtain values very close to whole numbers (e.g., 1.33 × 3 ≈ 4). Round to the nearest whole number.
• Write the Empirical Formula: Use these whole numbers as subscripts to construct the formula.

Table 1: Example Calculation for a Hypothetical Biomolecule

Element	Mass (%)	Mass in 100g (g)	Molar Mass (g/mol)	Moles	÷ by Smallest (3.33)	Ratio	Whole Number Ratio
Carbon (C)	40.0	40.0	12.01	3.33	1.00	1	1
Hydrogen (H)	6.7	6.7	1.01	6.63	1.99	2	2
Oxygen (O)	53.3	53.3	16.00	3.33	1.00	1	1

Resulting Empirical Formula: CH₂O [10].

Protocol 2: Deriving a Representative Empirical Formula for Biomass

Biomass is a complex, heterogeneous mixture of structural polymers (cellulose, hemicellulose, lignin), and other components. Its "empirical formula" is a weighted average representing the overall elemental ratio of the sample. The U.S. National Renewable Energy Laboratory (NREL) has established standardized Laboratory Analytical Procedures (LAPs) for this purpose [12].

Workflow Overview:

Detailed Procedure:

• Sample Preparation:

  • Dry the biomass sample in a convection oven or using an infrared moisture analyzer to determine the dry weight and moisture content [12].
  • Mill the biomass to pass through a 2-mm screen to ensure a uniform, representative particle size [12].

• Determination of Extractives:

  • Perform a solvent extraction (e.g., with water and ethanol) to remove non-structural components like sugars, oils, and pigments [12].
  • This step is critical for reporting the final composition on an "as-received" basis. The mass loss corresponds to the extractives fraction.

• Determination of Structural Carbohydrates and Lignin (Core Protocol):

  • This procedure quantifies the main polymeric components [12].
  • Two-Stage Acid Hydrolysis:
    • Primary Hydrolysis: React the extractives-free biomass with 72% sulfuric acid (H₂SO₄) at 30°C for 1 hour, with continuous stirring. This step solubilizes the polymeric carbohydrates.
    • Secondary Hydrolysis: Dilute the acid to 4% and hydrolyze in an autoclave at elevated temperature (e.g., 121°C). This step completes the breakdown of oligomers into monomeric sugars.
  • Analysis of Hydrolysate (Liquid Fraction):
    • Use High-Performance Liquid Chromatography (HPLC) to quantify the monomeric sugars (glucose, xylose, arabinose, etc.) in the liquid. These values are converted back to their anhydrous polymer masses (e.g., glucan, xylan) using anhydro corrections [12].
    • The hydrolysate is also analyzed for acid-soluble lignin (via UV-Vis spectroscopy) and carbohydrate degradation products like furfural.
  • Analysis of Solid Residue (Solid Fraction):
    • The remaining solid is filtered, dried, and weighed to determine acid-insoluble lignin (Klason lignin) [12].
    • The solid is then combusted at 575°C±25°C to determine the ash content, which is subtracted from the residue mass for an accurate lignin measurement [12].

• Summative Mass Closure:

  • The final composition is the sum of the masses of:
    • Structural Carbohydrates (Glucan, Xylan, etc.)
    • Acid-Insoluble Lignin
    • Acid-Soluble Lignin
    • Extractives
    • Ash
  • The total should approach 100% of the dry weight of the original sample, providing a mass closure that validates the analysis [12].
  • From the detailed composition, a representative empirical formula (e.g., CH~1.4~O~0.6~N~0.05~) can be calculated by aggregating the elemental contributions from each quantified component.

Table 2: Key Research Reagent Solutions for Biomass Analysis

Reagent/Item	Function in Protocol
Sulfuric Acid (H₂SO₄)	Primary catalyst for the two-stage hydrolysis process that breaks down structural polymers into quantifiable monomers [12].
HPLC System with Refractive Index (RI) / UV Detector	Quantifies concentrations of monomeric sugars (glucose, xylose), sugar alcohols, and degradation products in the hydrolysate [12].
De-ashing Cartridges	Used during HPLC sample preparation to remove salts that can interfere with the RI detector and produce false signals [12].
Vacuum Filtration Apparatus	Used with a defined crucible to separate the solid residue (acid-insoluble lignin) from the liquid hydrolysate after the second-stage hydrolysis [12].
Reference Biomass Materials (e.g., from NIST)	Homogeneous standard materials used to validate analytical methods and ensure accuracy and precision across measurements [12].

Application in BLSS Stoichiometric Modeling

In a BLSS, the mass flows of elements must be balanced to sustain the crew. A recent stoichiometric model of a fully closed MELiSSA (Micro-Ecological Life Support System Alternative) loop, designed for a crew of six, uses fixed chemical equations to describe the cycling of C, H, O, and N through its five compartments [1]. These equations rely on the empirical compositions of all inputs and outputs, from human waste to plant biomass.

For instance, the consumption of plant food by the crew (Compartment 5) and the subsequent processing of solid waste in bioreactors (Compartments 1-3) are modeled using stoichiometric equations. The model's high degree of closure—where most compounds exhibit zero loss between cycles—is predicated on accurate empirical formulas for these biomass streams [1]. Using an inaccurate empirical formula for, say, Limnospira (spirulina) biomass grown in Compartment 4a would lead to erroneous predictions of oxygen production or carbon dioxide consumption, ultimately jeopardizing the system's balance. Therefore, the rigorous analytical protocols described herein are not just best practices but necessities for viable BLSS design.

Defining the empirical formulas of biomolecules and biomass through standardized analytical protocols provides the non-negotiable data foundation for stoichiometric modeling of BLSS. The calculative method for pure compounds and the detailed, multi-step wet chemical procedures for heterogeneous biomass, as established by bodies like NREL, ensure the accuracy and reliability of this data. Integrating these precisely defined elemental ratios into mass flow models, such as those for the MELiSSA loop, is the key to achieving the high degree of material closure required for autonomous, long-duration human space exploration. This approach transforms the abstract concept of "biomass" into a quantifiable and manageable variable within a closed ecological system.

The Challenge of Achieving Full Closure in Bioregenerative Systems

For long-duration space missions beyond Earth's orbit, Bioregenerative Life Support Systems (BLSS) become essential for human survival by regenerating resources through biological processes. The central challenge lies in achieving full material closure—creating a system where waste products are entirely recycled into food, water, and oxygen without significant resupply from Earth [13] [3]. On missions exceeding three months, the equivalent system mass (ESM) trade-offs favor BLSS over purely physiochemical systems due to reduced resupply requirements [14]. Stoichiometric modeling provides the foundational framework for understanding and balancing the mass flows of carbon, hydrogen, oxygen, and nitrogen through all system compartments, enabling the design of a truly closed ecosystem [1]. This application note details the specific challenges and protocols for achieving this closure, with a focus on quantitative mass balance and experimental validation.

Key Obstacles to System Closure

Mass Flow Imbalances and Cycling Acceleration

In a materially closed BLSS, the small reservoir sizes of critical elements, compared to Earth's biosphere, lead to accelerated cycling rates and heightened sensitivity to imbalances [13]. A transient disruption in one process can rapidly propagate through the entire system, causing destabilizing fluctuations in oxygen, carbon dioxide, or nutrient levels. Accumulation of trace gases or recalcitrant materials not accounted for in the initial stoichiometric model can further jeopardize closure by tying up elements in unusable forms [13]. These challenges necessitate robust, resilient biological communities and precise control systems not required in terrestrial or resupply-dependent environments.

Integration and Operational Hurdles

• Nitrogen Recovery Complexity: Nitrogen is a critical element for protein synthesis in food. While urine is the primary source of recoverable nitrogen (85%), its efficient conversion into a plant-accessible form (e.g., nitrate) via processes like nitrification is a complex, multi-step biological challenge [15]. Inadequate nitrogen recovery directly impacts food production and system closure.
• Food Production Limitations: Higher plants are vital as food producers and for air and water regeneration. However, for full closure, staple crops (e.g., wheat, potato) with long growth cycles must be incorporated, requiring significant growing area and precise resource allocation [3]. The choice between fast-growing "salad machine" species and calorie-dense staple crops directly impacts the system's mass balance.
• Pathogen and Pest Management: The introduction of plants and their associated microbiomes creates the risk of phytopathogen outbreaks, as demonstrated by a Fusarium oxysporum outbreak in the International Space Station's Veggie module [14]. Effective Integrated Pest Management (IPM) protocols are therefore not optional but essential for maintaining the stability and productivity of the plant compartment [14].

Quantitative Analysis of Mass Flows

Elemental Mass Flow in a Conceptual BLSS

Stoichiometric modeling tracks the flow of elements through interconnected compartments. The following table summarizes the steady-state mass flow for a crew of six, based on a MELiSSA-inspired model aiming for high closure [1].

Table 1: Daily Elemental Mass Flow for a Crew of Six in a Closed BLSS (Data adapted from [1])

Element	Human Consumption (g/day)	Plant & Algal Uptake (g/day)	Waste Processing Output (g/day)	Closure Efficiency
Carbon (C)	811.2	810.5	810.5	~99.9%
Hydrogen (H)	111.6	111.4	111.4	~99.8%
Oxygen (O)	1062.4	1060.1	1060.1	~99.8%
Nitrogen (N)	19.8	19.7	19.7	~99.5%

System Performance Metrics

The performance of BLSS ground demonstrators can be evaluated against key closure metrics.

Table 2: Performance Metrics for BLSS Closure

Performance Metric	Target for Full Closure	Current State-of-the-Art (Example)
Food Production Closure	100%	Varies; often only a fraction is targeted in tests [1]
Oxygen Closure	100%	High (>99%) achievable in models [1]
Water Recovery	>98%	~96.5% on ISS (physicochemical); targets are higher for BLSS [15]
Nitrogen Recovery	>98%	A major focus of current R&D (e.g., MELiSSA C3) [15]
Mission Duration without Resupply	>3 years	1-year demonstration in Lunar Palace 1 [2]

Experimental Protocols for Closure Research

Protocol: Stoichiometric Model Development for BLSS

This protocol outlines the creation of a stoichiometric model to describe mass flows in a BLSS.

1. Research Question: How do the elements C, H, O, and N cycle through all compartments of a BLSS at steady state, and what are the system's closure points?

2. Experimental Workflow:

3. Procedure:

• Step 1: System Definition. Define the functional compartments of the BLSS loop (e.g., C1: Waste Bioreactor, C2: Photoheterotrophs, C3: Nitrifier, C4a: Algae, C4b: Higher Plants, C5: Crew) [1].
• Step 2: Stream Identification. List all material inputs and outputs for each compartment, including human waste, plant biomass, food, water, and gases [1] [16].
• Step 3: Formula Assignment. Assign empirical chemical formulas to all components. For example, human solid waste can be represented as CH1.8O0.5N0.1, and algal biomass as CH1.8O0.4N0.2 [1] [17].
• Step 4: Equation Balancing. Write balanced chemical equations for the key processes in each compartment (e.g., nitrification, photosynthesis, human consumption). Ensure mass balance for each element [1] [18].
• Step 5: Model Implementation. Implement the system of equations in a spreadsheet or computational solver. Use a crew of six as a baseline for scaling input and output fluxes [1].
• Step 6: Iteration and Validation. Run the model iteratively to balance all flows. Validate the model outputs against data from experimental ground-based facilities like the MELiSSA Pilot Plant or Lunar Palace [1] [2].

4. Interpretation: A successful model will show minimal losses for most compounds at steady state, with oxygen and CO2 exhibiting only minor losses between iterations [1]. Non-closing fluxes indicate where system inefficiencies lie or where additional processing compartments are required.

Protocol: Operation of a Nitrifying Bioreactor for Nitrogen Recovery

This protocol details the operation of the nitrifying compartment (e.g., MELiSSA C3), which is critical for converting ammonia from waste streams into nitrate fertilizer for plants [15].

1. Research Question: Can a nitrifying bioreactor stably convert the ammonium load from a crew's urine into nitrate with a conversion efficiency of >98%?

2. Experimental Workflow:

3. Procedure:

• Step 1: Feedstock Preparation. Collect and chemically stabilize pretreated urine to prevent scaling and urea hydrolysis. Acidification (e.g., with H₃PO₄) converts volatile ammonia to non-volatile ammonium and reduces scaling [15].
• Step 2: Inoculation and Operation. Inoculate the bioreactor with a defined nitrifying consortium (e.g., Nitrosomonas europaea, Nitrobacter winogradskyi). Operate the reactor in a continuous or fed-batch mode.
• Step 3: Process Monitoring. Continuously monitor key parameters: Ammonium (NH₄⁺) and Nitrite (NO₂⁻) concentrations (target: near zero), Nitrate (NO₃⁻) concentration (target: steady increase proportional to NH₄⁺ decrease), and pH (controlled around 7.5-8.0 for optimal nitrifier activity) [15].
• Step 4: Product Formulation. The effluent, rich in nitrate, is combined with other recovered nutrients (P, K) and trace elements to form a complete hydroponic nutrient solution.
• Step 5: Plant Growth Trial. Deliver the nutrient solution to a plant growth compartment (e.g., growing lettuce or wheat). Monitor plant biomass production and nitrogen uptake efficiency [3].

4. Interpretation: Successful nitrogen closure is demonstrated by a high conversion efficiency of urine nitrogen into plant biomass nitrogen, with minimal accumulation of intermediate nitrite or gaseous nitrogen losses.

The Scientist's Toolkit: Key Reagents and Materials

Table 3: Essential Research Reagents for BLSS Stoichiometry and Closure Experiments

Reagent/Material	Function in BLSS Research
Defined Microbial Consortia (e.g., Nitrosomonas, Nitrobacter)	To conduct specific waste recycling processes like nitrification with predictable stoichiometry [15].
Hydroponic Nutrient Solutions	To provide precise mineral nutrition for plant growth studies and validate nutrient uptake models [3].
Chemical Tracers (e.g., ¹⁵N-labeled Urea)	To quantitatively track the fate of nitrogen atoms through different compartments (e.g., urine to plant biomass) [15].
Standardized Synthetic Waste Feeds	To simulate human waste (feces, urine) with a consistent chemical composition for reproducible bioreactor experiments [1] [16].
Gas Analysis Standards (e.g., CO2, O2, CH4)	To calibrate sensors for real-time monitoring of atmospheric gases, crucial for detecting leaks or imbalances [13].

Achieving full closure in Bioregenerative Life Support Systems remains a formidable challenge that hinges on precise stoichiometric balancing of mass flows and the robust integration of biological components. While current research, exemplified by projects like MELiSSA and the Lunar Palace, has demonstrated the feasibility of long-duration operation, gaps in nitrogen recovery, trace gas management, and system stability under space conditions persist. Future work must focus on closing these specific loops through advanced modeling, ground-based testing in integrated facilities, and the development of dynamic control strategies that can respond to the inherent variability of biological systems. Success will enable the sustainable human exploration of deep space.

Computational Frameworks: From FBA to Genome-Scale Models

Flux Balance Analysis (FBA) is a mathematical computational approach for analyzing the flow of metabolites through metabolic networks. It calculates the steady-state fluxes in a biochemical reaction network to predict outcomes like growth rate or metabolite production. This methodology is particularly valuable for simulating complex systems such as Bioregenerative Life Support Systems (BLSS), where understanding mass flows of elements like carbon, hydrogen, oxygen, and nitrogen is critical for sustainability [1].

FBA operates on constraint-based modeling, using the stoichiometric coefficients of every reaction in a genome-scale metabolic model (GEM) to form a numerical matrix [19]. A GEM contains all known metabolic reactions for an organism and the genes encoding each enzyme [20]. The core principle involves defining a solution space bounded by constraints and applying an optimization function to identify a flux distribution that maximizes a biological objective, such as biomass production [21] [19].

For BLSS research, which aims to create materially closed loops for long-duration space missions, FBA provides a powerful in-silico tool. It enables scientists to model and optimize the metabolic interactions between crew and organisms (e.g., plants, microalgae, bacteria) that recycle waste into oxygen, water, and food [1]. By predicting how these systems consume and regenerate resources, FBA can inform the design of more robust and efficient BLSS, helping to achieve the high degree of closure necessary for mission autonomy [1].

Core Principles and Key Assumptions

The application of FBA is founded on several key principles and assumptions:

• Steady-State Assumption: The model assumes the network is in a steady state, meaning that for each internal metabolite, the rate of production equals the rate of consumption, resulting in no net accumulation [19].
• Stoichiometric Constraints: The stoichiometric matrix (S), derived from the metabolic reconstruction, defines the mass balance constraints under which the system operates [19].
• Physiological Constraints: Additional constraints, based on experimental measurements, are applied to flux variables, such as substrate uptake rates or the maximum capacity of certain reactions [20].
• Objective Function: An objective function is chosen to represent a biological goal, which the model optimizes to find a particular flux distribution within the solution space. A common objective is the Biomass Objective Function (BOF), which simulates the production of all biomass precursors in the correct proportions to support cellular growth [20].

Protocol: A Step-by-Step Guide to Performing FBA

The following protocol outlines the standard workflow for conducting an FBA.

Step 1: Define the Metabolic Network and Stoichiometric Matrix

• Action: Compile all metabolic reactions relevant to the organism and system under study into a stoichiometric matrix. For genome-scale models, this can involve thousands of reactions and metabolites [20].
• BLSS Context: In a multi-compartment BLSS like MELiSSA, this may require integrating models for the different organisms inhabiting each compartment to describe the cycling of elements [1].

Step 2: Set Constraints on the Reaction Network

• Action: Apply constraints to the flux variables (v). These typically include:
  • Reversibility Constraints: Define whether each reaction can proceed in both forward and reverse directions.
  • Capacity Constraints: Set upper and lower bounds (vmax, vmin) for reaction fluxes based on physiological data [19].
• Example: In a study simulating E. coli growth, the uptake rate of glucose might be constrained to a measured value [19].

Step 3: Formulate the Biomass Objective Function (BOF)

• Action: Define the BOF as a pseudo-reaction that consumes all necessary biomass precursors (e.g., amino acids, nucleotides, lipids) in their known cellular proportions [20].
• Detail: The BOF can be formulated at different levels:
  • Basic: Accounts for the macromolecular composition of the cell [20].
  • Intermediate: Includes biosynthetic energy requirements (e.g., ATP cost for polymerization) [20].
  • Advanced: Incorporates vitamins, cofactors, and data from genetic mutants to define a "core" essential biomass [20].

Step 4: Solve the Linear Programming Problem

• Action: Perform the optimization. The standard formulation is:
  • Maximize Z = c^T v
  • Subject to: S • v = 0
  • and vmin ≤ v ≤ vmax
• where Z is the objective function (e.g., biomass growth rate), c is a vector of weights indicating how much each flux contributes to the objective, S is the stoichiometric matrix, and v is the flux vector [19] [20].

Step 5: Analyze and Validate the Results

• Action: Interpret the predicted flux distribution. Compare the in-silico predictions, such as growth rates or substrate uptake/secretion rates, with experimental data to validate the model [20].

Application Notes: FBA for BLSS Stoichiometric Modeling

Integrating FBA into a Multi-Compartment BLSS Model

In the MELiSSA BLSS concept, mass flows connect several compartments. FBA can model each biological compartment (e.g., nitrifying bacteria, photoheterotrophic bacteria, higher plants) as an individual metabolic network [1]. The challenge is to appropriately define the exchange fluxes between compartments, ensuring that the waste outputs from the crew (C5) become the inputs for waste-processing compartments (C1, C2, C3), and that the nutrients produced by these compartments support the growth of autotrophic organisms (C4) that, in turn, sustain the crew [1].

Refining Predictions with Enzyme Constraints

A known limitation of traditional FBA is that it can predict unrealistically high fluxes. This can be addressed by incorporating enzyme constraints using workflows like ECMpy [19]. This method caps the flux through a reaction based on enzyme availability and catalytic efficiency (Kcat), adding a layer of biophysical reality without altering the stoichiometric matrix. For a BLSS model, this increases the accuracy of predicting how genetic modifications in key organisms might affect overall system flux.

Lexicographic Optimization for Realistic Scenarios

Optimizing for a single objective, like L-cysteine export in an engineered organism, can result in solutions with zero biomass growth, which is biologically unrealistic [19]. Lexicographic optimization is a solution: the model is first optimized for biomass. It is then constrained to require a percentage of that maximum growth (e.g., 30-90%) while a second objective (e.g., product synthesis) is optimized [19]. This ensures the solution reflects a growing, metabolically active system, which is essential for a sustainable BLSS.

Data Presentation

Table 1: Example Modifications to a Base Metabolic Model for FBA

This table illustrates how a base model like E. coli's iML1515 can be modified to reflect genetic engineering and simulate altered metabolic behavior, a key process for optimizing BLSS organisms [19].

Parameter	Gene/Enzyme/Reaction	Original Value	Modified Value	Justification
Kcat_forward	PGCD	20 1/s	2000 1/s	100-fold increase in mutant enzyme activity [19]
Kcat_forward	SERAT	38 1/s	101.46 1/s	Removal of feedback inhibition [19]
Gene Abundance	SerA/b2913	626 ppm	5,643,000 ppm	Reflects modified promoter and copy number [19]

Table 2: Uptake Reaction Bounds for a Defined Growth Medium (SM1 + LB)

FBA simulations require defining the environmental conditions through constraints on uptake reactions. This table provides an example for a specific medium [19].

Medium Component	Associated Uptake Reaction	Upper Bound (mmol/gDW/h)
Glucose	EXglcDe_reverse	55.51
Ammonium Ion	EXnh4e_reverse	554.32
Phosphate	EXpie_reverse	157.94
Sulfate	EXso4e_reverse	5.75
Thiosulfate	EXtsule_reverse	44.60

The Scientist's Toolkit: Essential Reagents and Software

Research Reagent Solutions

• Genome-Scale Metabolic Model (GEM): A structured database of all known metabolic reactions for an organism (e.g., iML1515 for E. coli K-12). It serves as the fundamental scaffold for constructing an FBA model [19].
• Stoichiometric Matrix: A mathematical representation of the metabolic network derived from the GEM. It defines the mass balance constraints for the FBA problem [19].
• Biomass Objective Function (BOF): A pseudo-reaction that defines the drain of metabolic precursors required for cell growth. It is the most common objective function used to simulate growth [20].
• Enzyme Constraint Data: Data on enzyme kinetic parameters (Kcat values) and abundances, used to add capacity constraints on reactions and improve the predictive accuracy of FBA [19].

Computational Tools

• COBRApy: A popular Python package for performing constraint-based reconstruction and analysis, including FBA [19].
• ECMpy: A workflow for incorporating enzyme constraints into existing GEMs without altering the stoichiometric matrix [19].

Constructing Genome-Scale Metabolic Models

Genome-Scale Metabolic Models (GSSMs) are computational reconstructions of the metabolic network of an organism, based on its genomic annotation. They represent a comprehensive, stoichiometric accounting of the reactions and metabolites that constitute metabolism, enabling in silico simulation of metabolic fluxes. In the context of Bioregenerative Life Support Systems (BLSS), GSSMs are indispensable tools for modeling mass flows of carbon, hydrogen, oxygen, and other elements, allowing researchers to predict how these closed-loop systems will behave under various conditions. The construction of a high-quality GSSM involves a multi-step process of network reconstruction, curation, and mathematical formulation, culminating in a model that can be used for simulation and analysis via techniques such as Flux Balance Analysis (FBA) [22] [23].

GSSM Reconstruction Workflow and Protocol

The construction of a GSSM is a systematic process that transforms genomic information into a mathematical model capable of predicting phenotypic behavior. The following workflow outlines the key stages, from initial data collection to final model validation.

Figure 1: A high-level workflow for the systematic reconstruction of a Genome-Scale Metabolic Model.

Protocol: Draft Model Reconstruction and Curation

This protocol details the steps for generating a draft model from a genome annotation and refining it into a functional GSSM.

• Inputs:
  • Annotated genome sequence (e.g., from RAST or Prokka).
  • Biochemical reaction database (e.g., ModelSEED, KEGG).
• Procedure:
  • Generate Draft Model: Use an automated reconstruction tool (e.g., the Build Metabolic Model app in KBase) to map annotated genes to reactions from a biochemical database. This process creates an initial draft network [23].
  • Define the Biomass Objective Function (BOF): Assemble a reaction that represents the composition of a new cell, including all necessary metabolites in their correct proportions (e.g., amino acids, nucleotides, lipids, cofactors). The BOF is typically set as the objective function for simulating growth [22].
  • Perform Gapfilling: The draft model will likely be unable to produce all biomass precursors due to missing annotations. The gapfilling process algorithmically adds a minimal set of reactions from a database to enable growth on a specified medium.
    • Algorithm: The process uses Linear Programming (LP) to minimize the sum of flux through the added (gapfilled) reactions. Transporters and non-KEGG reactions are assigned higher penalties to favor biologically plausible solutions [23].
    • Media Selection: It is recommended to perform initial gapfilling on a minimal medium. This ensures the algorithm adds the maximal set of reactions required for the biosynthesis of essential substrates. Gapfilling on "complete" media (where all transportable compounds are available) can be performed subsequently, but will result in a model that is more dependent on nutrient uptake [23].
  • Validate the Model: Simulate growth under different environmental conditions (e.g., varying carbon sources) using Flux Balance Analysis (FBA). Compare the predicted growth outcomes, substrate uptake rates, and by-product secretion rates against experimental data from the literature to assess model accuracy [22].

Key Reagents and Computational Tools for GSSM Construction

The following table details essential data resources and tools required for the construction and analysis of GSSMs.

Table 1: Key Research Reagent Solutions for GSSM Construction

Item Name	Function/Application	Critical Specifications
AGORA2 Resource	A database of curated, strain-level GEMs for 7,302 human gut microbes. Used as a source of pre-reconstructed models or as a reference for reaction and metabolite formatting [22].	Includes models with mapped taxonomic and phenotypic data.
ModelSEED Biochemistry	A standardized biochemistry database that provides controlled vocabularies for roles, reactions, and compounds. Essential for ensuring consistency during draft model generation [23].	Integrated into the KBase reconstruction pipeline.
KBase Gapfill App	A computational tool that identifies and adds a minimal set of reactions to a draft model to enable it to produce biomass on a specified growth medium [23].	Uses Linear Programming (LP) with a cost function for reactions; allows user-defined media conditions.
Custom Media Formulation	A defined set of extracellular metabolites available to the model during simulation and gapfilling. Critical for contextualizing the model to a specific environment, such as a BLSS [23].	Can be minimal or complex; must be defined in a compatible format (e.g., in KBase).
RAST Annotation Pipeline	A service for annotating genome sequences. Its functional roles are recommended for metabolic modeling in KBase due to their use as a controlled vocabulary for deriving reactions [23].	Provides consistent gene-to-function assignments.

Advanced Applications and Customization in a BLSS Context

Once a functional GSSM is constructed, it can be used for advanced in silico analyses to probe system capabilities and design strategies for optimizing BLSS mass flows.

Protocol: Simulating Interspecies Interactions in a BLSS

A primary application of GSSMs in a BLSS is modeling the metabolic exchange between organisms (e.g., plants, microbes, and humans) to stabilize the closed-loop system.

• Objective: To predict the outcome of introducing a microbial LBP (Live Biotherapeutic Product) or other biological component on the existing BLSS community.
• Procedure:
  • Select Candidate Models: Obtain GEMs for the resident BLSS microbes and the candidate organism (e.g., from AGORA2 or by constructing them de novo) [22].
  • Define the Community Medium: Formulate a custom media condition that represents the BLSS nutrient environment.
  • Simulate Pairwise Interactions: For the candidate strain and each key resident microbe, perform a pairwise simulation.
    • A. Maximize the growth of the candidate strain.
    • B. Take the fermentative by-products secreted by the candidate and add them as nutritional inputs to the model of the resident microbe.
    • C. Simulate the growth of the resident microbe with and without these candidate-derived metabolites.
  • Analyze Results: Compare the growth rates from step 3C. An increase suggests a commensal or mutualistic relationship (e.g., cross-feeding), while a decrease suggests competition or inhibition. This helps screen for candidates that positively influence the BLSS community [22].

Customization for BLSS: Visualizing Regulatory Interactions

Understanding dynamic regulation is key for BLSS management. The concept of Regulatory Strength (RS) can be used to visualize how metabolite pools regulate reaction fluxes within a GSSM. The RS quantifies the strength of an effector (inhibitor or activator) on a reaction step at a given metabolic state, providing a percentage that indicates its contribution to the total regulation of that reaction [24]. The following diagram illustrates the logic for determining and interpreting these interactions.

Figure 2: A logic flow for calculating and interpreting Regulatory Strength (RS) to visualize metabolite-reaction interactions in a GSSM [24].

Integrating Omics Data into Stoichiometric Network Models

The integration of multi-omics data into stoichiometric network models represents a transformative approach in systems biology, enabling researchers to bridge the gap between genomic potential and observed phenotypic behavior. This integration is particularly critical for complex biological systems such as Bioregenerative Life Support Systems (BLSS), where understanding and predicting mass flows of elements like carbon, hydrogen, oxygen, and nitrogen is essential for system stability and closure [1]. Stoichiometric models, traditionally based on Genome-Scale Metabolic Models (GEMs), provide a structured framework for analyzing the organization and dynamics of cellular mechanisms but often lack the capacity to incorporate real-time molecular data [25] [26]. The advent of high-throughput omics technologies—including genomics, transcriptomics, proteomics, and metabolomics—has generated unprecedented amounts of data that, when effectively integrated, can constrain and refine these models, significantly enhancing their predictive accuracy [27] [28].

For BLSS research, which aims to create sustainable closed-loop environments for long-duration space missions, this integration is paramount. These systems rely on interconnected compartments of organisms to recycle waste into oxygen, water, and food [1]. The precise quantification of mass flows through metabolic networks is thus crucial for achieving a high coefficient of closure—the percentage of resources regenerated within the system [29]. This protocol details methodologies for embedding multi-omics data into stoichiometric models to achieve such precision, providing a structured guide for researchers and scientists engaged in predictive metabolic modeling.

Background and Principles

Stoichiometric Modeling Fundamentals

Stoichiometric models are built around the concept of mass balance and the stoichiometry of biochemical reactions within a metabolic network. The core mathematical framework is often based on Constraint-Based Reconstruction and Analysis (COBRA), which assumes steady-state metabolite concentrations. This is represented as:

S · v = 0

where S is the stoichiometric matrix (m × n), with m metabolites and n reactions, and v is the flux vector of reaction rates [26]. The solution space is constrained by enzyme capacity and nutrient availability, typically leading to a linear programming problem where an objective function (e.g., biomass production) is maximized or minimized.

The integration of omics data introduces additional constraints that refine this solution space. For instance, proteomic data can be used to constrain the maximum flux through a reaction based on the measured abundance of its catalyzing enzyme and its turnover number [26]. This moves the model from a genetically defined potential state to a context-specific state that reflects actual physiological conditions.

Omics Data Types and Their Roles

Different omics layers provide distinct and complementary information for constraining metabolic models:

• Genomics defines the network structure itself, providing the gene-protein-reaction (GPR) associations that form the foundation of GEMs [30].
• Transcriptomics indicates which genes are being actively transcribed, often used as a proxy for enzyme capacity though subject to post-transcriptional regulation [27].
• Proteomics directly quantifies enzyme abundance, offering a more reliable constraint for flux calculations [26].
• Metabolomics provides snapshots of metabolite pool sizes, which can inform on thermodynamic constraints and reaction directions [27].

In the context of BLSS, the primary goal is to model the cycling of key elements (C, H, O, N) through the system's compartments. A successfully integrated model can predict how perturbations in one compartment affect the entire system, which is vital for managing essential outputs like oxygen and food production [1].

Computational Approaches and Methodologies

Several computational approaches have been developed to integrate omics data into stoichiometric models, which can be broadly categorized into four main strategies [26]. Table 1 summarizes these approaches, their characteristics, and representative algorithms.

Table 1: Categories of Methods for Integrating Omics Data into Stoichiometric Models

Category	Description	Key Methods	Data Requirements	Output
Proteomics-Driven Flux Constraints	Uses enzyme abundance data to directly constrain upper flux bounds.	FBAwMC [26], MOMENT [26]	Quantitative proteomics, enzyme turnover numbers	Context-specific flux distributions
Proteomics-Enriched Stoichiometric Matrix Expansion	Expands the stoichiometric matrix to include explicit reactions for protein synthesis and degradation.	GECKO [26]	Proteomics, enzyme kinetic parameters	Resource allocation-aware flux solutions
Proteomics-Driven Flux Estimation	Uses statistical methods to integrate expression data and map it onto the network.	IOMA [26], MADE [26]	Relative or absolute proteomics/transcriptomics	Condition-specific metabolic states
Fine-Grained Methods	Incorporates detailed transcriptional and translational processes.	ETFL [26]	Multi-omics data (mRNA, protein, flux)	Integrated predictions of mRNA, enzyme, and flux

Hybrid Machine Learning Approaches

Recent advances have introduced hybrid frameworks that combine mechanistic stoichiometric models with data-driven machine learning (ML). The Metabolic-Informed Neural Network (MINN) is one such architecture that embeds GEMs within a neural network, allowing for the seamless integration of multi-omics data to predict metabolic fluxes [25]. This approach leverages the pattern recognition strength of ML while respecting the biochemical constraints enforced by the stoichiometric model. Similarly, NEXT-FBA represents another hybrid stoichiometric/data-driven approach designed to improve intracellular flux predictions [31].

These hybrid models are particularly valuable for addressing the "omics cascade"—the sequential flow of information from genes to transcripts, proteins, and metabolites—which is influenced by numerous regulatory mechanisms and environmental factors [27]. By learning complex, non-linear relationships from data while adhering to stoichiometric constraints, they can achieve higher predictive accuracy than purely mechanistic or purely data-driven approaches alone.

Application Notes and Protocols

Protocol: Integrating Proteomics Data into a BLSS Stoichiometric Model

This protocol details the steps for integrating quantitative proteomics data into a stoichiometric model of a BLSS compartment, using the GECKO (GEnome-scale model with Enzymatic Constraints using Kinetic and Omics data) method as a framework [26].

Experimental Design and Workflow

The entire process, from model preparation to simulation and validation, is outlined in Figure 1 below.

Figure 1: Workflow for integrating proteomics data into a stoichiometric model using a GECKO-like approach.

Step-by-Step Procedure

Step 1: Model and Data Preparation

• GEM Preparation: Obtain a high-quality Genome-Scale Metabolic Model (GEM) for the organism of interest (e.g., Limnospira indica or a higher plant from the BLSS C4 compartment) [1] [26].
• Proteomics Acquisition: Generate or acquire quantitative proteomics data for the target organism under the specific BLSS growth condition (e.g., using mass spectrometry). Data should be in units of mg protein per gDW (gram dry weight) of cells [26].

Step 2: Model Expansion with Enzymatic Constraints

• Expand Stoichiometric Matrix: The GEM is expanded to include pseudo-reactions that represent the investment of enzymes in catalytic steps. This links metabolic fluxes to enzyme usage.
• The enzyme allocation constraint is formulated as: ∑ (vi / kcati) ≤ Ptot where v_i is the flux of reaction i, k_cat_i is the turnover number of the enzyme catalyzing reaction i, and P_tot is the total enzyme pool capacity [26].

Step 3: Incorporation of Proteomics Data

• Apply Measured Enzyme Levels: Use the proteomics data to constrain the upper bound for each enzyme's usage. For an enzyme E_i measured at a concentration [Ei], the flux through its catalyzed reaction is constrained by: vi ≤ [Ei] × kcat_i
• For enzymes with missing measurements, apply a default constraint or use gap-filling algorithms [26].

Step 4: Simulation and Analysis

• Define Objective: Set an appropriate biological objective function. While biomass maximization is common, for BLSS, objectives like oxygen production (C4 compartment) or waste degradation (C1 compartment) may be more relevant [1].
• Perform Flux Analysis: Solve the constrained model using Parsimonious Flux Balance Analysis (pFBA) or similar techniques to obtain a flux distribution that satisfies the proteomic constraints and optimizes the objective [26] [31].
• Validate Predictions: Compare predicted fluxes (e.g., substrate uptake, production rates) with experimentally measured rates, if available. For BLSS, this could involve comparing predicted CO2 uptake and O2 production by algae or plants to measured gas exchange data [1].

Reagent and Computational Tools

Successful implementation of this protocol requires specific reagents and computational tools. Table 2 lists the essential components of the "Researcher's Toolkit" for this workflow.

Table 2: Research Reagent and Computational Solutions for Omics-Stoichiometric Integration

Category	Item	Specifications / Function	Example Use in Protocol
Wet-Lab Reagents	Protein Lysis Buffer	For efficient cell disruption and protein extraction from BLSS organism samples (e.g., microalgae, higher plants).	Preparing samples for mass spectrometry-based proteomics.
	Quantitative Proteomics Kit (e.g., TMT/iTRAQ)	For isobaric labeling of peptides to enable multiplexed, relative quantification of protein abundance across different BLSS conditions.	Comparing enzyme abundance between different BLSS operational stages.
	Internal Standard (e.g., SILAC)	Labeled amino acids for spike-in absolute protein quantification.	Determining absolute enzyme concentrations (mg/gDW).
Software & Databases	COBRA Toolbox	A MATLAB toolbox for constraint-based modeling. The GECKO toolbox is built upon it.	Implementing the model expansion and simulation steps [26].
	R/Python Environment	For data pre-processing, statistical analysis, and visualization of omics data.	Normalizing proteomics data and generating correlation plots.
	Genome-Scale Model (GEM) Database (e.g., BiGG Models)	A repository of curated GEMs for various organisms.	Sourcing a starting GEM for a BLSS-relevant organism [26].
	Turnover Number (k_cat) Database (e.g., SABIO-RM)	A database of enzyme kinetic parameters.	Retrieving k_cat values for enzymatic constraints [26].

Data Analysis and Interpretation

Handling Multi-Omic Data Challenges

Integrating multiple omics datasets introduces challenges such as data heterogeneity, missing values, and different scales of measurement. A pre-processing pipeline is essential [27]:

• Normalization: Normalize omics data to account for technical variation (e.g., using quantile normalization for transcriptomics or proteomics).
• Imputation: Use appropriate methods (e.g., k-nearest neighbors, matrix factorization) to handle missing values in proteomics or metabolomics data, being cautious not to introduce bias.
• Data Scaling: Ensure all data types are on comparable scales before integration, for instance, by converting to Z-scores.

For BLSS applications, it is critical to align the omics sampling time-points with the steady-state assumption of the stoichiometric model. Systems like MELiSSA are often modeled at steady state, so omics data should be collected after the system has reached a stable operational point [1].

Validation of Integrated Models

Model validation is a critical step. Several approaches can be employed:

• Internal Validation: Compare predicted fluxes of exchange reactions (e.g., CO2 consumption, O2 production) against measured rates not used to constrain the model [1] [26].
• External Validation: Use the model to predict the system's behavior under a new condition (e.g., a different light regime in the photobioreactor) and test these predictions experimentally.
• Comparison to Benchmarks: Compare the performance of the integrated model (e.g., GECKO) against a baseline model (e.g., standard FBA) in terms of flux prediction accuracy [26]. A successful integration should yield predictions that are closer to experimental observations.

In the context of BLSS closure, a key validation metric is the accurate prediction of mass flow rates for carbon, hydrogen, oxygen, and nitrogen between compartments. The model should be able to simulate a high degree of closure, with minimal losses of these key elements [1] [29].

Troubleshooting and Best Practices

Common Issues and Solutions

• Issue 1: Model Infeasibility after adding omics constraints.
  • Solution: Check for overly restrictive bounds. Relax constraints for reactions with low-confidence proteomics measurements or estimated k_cat values. Ensure the biomass reaction is appropriately curated for the BLSS organism [26].
• Issue 2: Large discrepancies between predicted and measured fluxes.
  • Solution: Verify the quality and quantification of the proteomics data. Re-assess the assigned kcat values, as these are often a major source of uncertainty. Consider using a range of kcat values in a sensitivity analysis [26].
• Issue 3: Incomplete Genome Annotation for non-model BLSS organisms.
  • Solution: Use comparative genomics tools to transfer annotations from well-annotated model organisms. Supplement the model with gap-filling techniques based on physiological data [1].

Recommendations for BLSS Research

• Compartment-Specific Modeling: BLSS like MELiSSA consist of multiple compartments (C1-C5). Develop and constrain separate models for each compartment (bacteria, algae, plants) before linking them through mass flow equations of exchanged compounds (CO2, O2, nutrients) [1].
• Dynamic Integration: For a more realistic simulation, consider moving from steady-state to dynamic hybrid modeling (e.g., NEXT-FBA, MINN) to capture how mass flows and omics profiles change over time in response to crew consumption and waste output [25] [31].
• Focus on Key Elements: Prioritize the accurate modeling of C, H, O, and N flows, as these are the most critical for life support. The stoichiometric equations should be balanced for these elements to ensure mass conservation [1] [29].

Bioregenerative Life Support Systems (BLSS) are pivotal for the future of long-duration space exploration, as they can significantly reduce mission mass and volume by closing the material loops of human metabolism [1] [32]. These systems use an artificial ecosystem of microorganisms, microalgae, and higher plants to break down human waste into nutrients and CO₂, which in turn provide food, oxygen, and fresh water for the crew [1]. The MELiSSA (Micro-Ecological Life Support System Alternative) project, led by the European Space Agency, is one of the most advanced BLSS concepts, designed as a five-compartment loop to achieve this cycling [1] [33]. For missions without resupply possibilities, a fully closed BLSS that generates all metabolic resources autonomously is essential [1]. Stoichiometric modeling, which is based on the mass balances of chemical elements, provides the mathematical framework to describe, simulate, and optimize the material flows in such a complex system [34] [35]. This case study details the application of a stoichiometric model for a MELiSSA-inspired BLSS designed to meet the metabolic needs of a crew of six.

Stoichiometric Modeling Fundamentals

Stoichiometric modeling of metabolic networks is a constraint-based approach that relies on the fundamental principle of mass conservation [34] [35].

Core Mathematical Principles

The system is defined by its stoichiometric matrix, denoted as S, where rows represent metabolites and columns represent biochemical reactions. The entry Sᵢⱼ is the stoichiometric coefficient of metabolite i in reaction j [34] [35]. The dynamics of the metabolite concentration vector x are governed by the differential equation: dx/dt = S ⋅ v where v is the flux vector of reaction rates [34]. At a metabolic steady state, the time derivative is zero, leading to the core equation for flux balance analysis: S ⋅ v = 0 [34] [36]. This equation represents a system of linear balance constraints, meaning that for each internal metabolite, the net rate of production must equal the net rate of consumption. The flux vector v is further constrained by lower and upper bounds (a ≤ v ≤ b) that encode thermodynamic irreversibility and enzyme capacity [36] [35].

Modeling Workflow and Pathway Analysis

The process begins with a network reconstruction, defining all metabolites and reactions [36]. The subsequent analyses can be grouped into methodologies for system analysis and for determining flux solutions [35]. Key techniques include:

• Metabolic Flux Analysis (MFA): Uses measured external fluxes to compute the internal flux map [35].
• Flux Balance Analysis (FBA): Uses linear programming to predict fluxes by assuming the network maximizes a biological objective (e.g., biomass growth) [34] [35].
• Network-Based Pathway Analysis: Identifies elementary flux modes—non-decomposable steady-state pathways that define the network's inherent capabilities [35].

The following diagram illustrates the typical workflow for developing and applying a stoichiometric model.

MELiSSA Loop Compartmentalization and Stoichiometry

The MELiSSA loop is structured into five functionally specialized compartments that sequentially process waste and regenerate resources [1] [33].

Compartment Functions and Mass Flows

• C1: Thermophilic Anaerobic Digestion. This compartment liquefies solid waste (feces, inedible biomass) through thermophilic anaerobic fermentation. It converts complex organic polymers into volatile fatty acids (VFAs), CO₂, and minerals [1].
• C2: Photoheterotrophic Compartment. Bacteria in this compartment use light as an energy source to oxidize the VFAs from C1, effectively removing reducing power from the system. This process also contributes to air revitalization [1].
• C3: Nitrifying Compartment. This compartment performs nitrification, converting ammonia (NH₃) from urine and mineralized waste into nitrate (NO₃⁻), which is a preferred nitrogen source for plants [1] [37].
• C4: Photoautotrophic Compartments. This stage comprises two units:
  • C4a: Microalgae (e.g., Limnospira indica). The microalgae consume the CO₂ from the crew and earlier compartments, producing oxygen and biomass for food [1].
  • C4b: Higher Plants. Edible plants are grown hydroponically, using the nitrate from C3 and CO₂ to produce the bulk of the food, oxygen, and clean water [1] [37].
• C5: Human Crew. The crew consumes the oxygen, water, and food produced by C4, generating solid and liquid waste, CO₂, and mineralized water, thereby closing the loop [1].

The integrated flow of mass and energy through these compartments is illustrated below.

Simplified Stoichiometric Equations

The following table presents a simplified set of stoichiometric equations that describe the core transformations in a fully closed MELiSSA loop. These equations consider the cycling of Carbon (C), Hydrogen (H), Oxygen (O), and Nitrogen (N) [1].

Table 1: Key Stoichiometric Equations for the MELiSSA Loop Compartments.

Compartment	Representative Stoichiometric Equation	Primary Function
C1	Complex Organics (e.g., C₆H₁₀O₅) + H₂O → CH₃COOH (Acetate) + CO₂ + H₂ + ...	Liquefaction of solid waste into VFAs.
C2	CH₃COOH + 2O₂ + Light → 2CO₂ + 2H₂O	Oxidation of VFAs, CO₂ production.
C3	2NH₃ + 3O₂ → 2NO₂⁻ + 2H⁺ + 2H₂O2NO₂⁻ + O₂ → 2NO₃⁻	Conversion of ammonia to nitrate.
C4a	1.6 CO₂ + 0.4 NO₃⁻ + 1.6 H⁺ + H₂O + Light → C₁.₆H₂.₇O₁.₁N₀.₄ (Algal Biomass) + 1.6 O₂	Air revitalization and food production.
C4b	x CO₂ + y NO₃⁻ + z H₂O + Light → CₐHₑOᵢNₖ (Plant Biomass) + O₂	Primary food, O₂, and water production.
C5	CₐHₑOᵢNₖ (Food) + O₂ → CO₂ + H₂O + Urea + Solid Waste	Consumption of resources, production of waste.

Model Implementation for a Crew of Six

Mass Flow Budget

Implementing the stoichiometric model for a specific crew size requires balancing the input and output of every element. The following table provides a simplified, quantitative overview of the daily mass flows for a crew of six, based on the balanced stoichiometry of the closed loop [1].

Table 2: Daily Mass Flow Budget for a Crew of Six in a Fully Closed BLSS (values are illustrative).

Parameter	Value	Notes
Crew Metabolic O₂ Demand	2100 g/day	Based on average human O₂ consumption.
Crew Metabolic CO₂ Production	2500 g/day	Based on average human CO₂ production.
Edible Biomass Requirement	~1200 g/day	Dry weight equivalent of food for 6 people.
C4a Photobioreactor Volume	~20 m³	Sizing for Limnospira to meet O₂/food share.
C4b Growth Chamber Area	~150 m²	Sizing for higher plants to meet primary food needs.
Nitrogen Input Requirement	~15 g/day	As NO₃⁻ for plant growth, derived from waste.
System Closure Efficiency	>99% for 12/14 key compounds	Achievable with balanced compartment sizing [1].

Protocol: Dynamic Stoichiometric Simulation

This protocol outlines the steps for setting up and running a dynamic simulation of the BLSS using a stoichiometric model in a tool like a spreadsheet or specialized software.

1. Objective: To simulate the mass flows of C, H, O, and N through the BLSS over time and verify the system's closure for a crew of six. 2. Materials:

• Stoichiometric matrix defining all reactions in Comps C1-C4.
• Crew metabolic data (C5 inputs/outputs).
• Simulation platform (e.g., Python with SciPy, MATLAB, or even an advanced spreadsheet).

3. Procedure:

• Step 1: Define the Stoichiometric Matrix. Construct the full matrix S where columns are the reactions from Table 1 and rows are all metabolites involved (e.g., CO₂, O₂, H₂O, Biomass_C4a, NO₃⁻, NH₃, VFAs) [1] [35].
• Step 2: Set Initial Conditions and Flux Bounds. Define the initial pool sizes for all metabolites. Set flux bounds (a, b) for all reactions. For example, set the lower bound for the "Human O₂ consumption" flux to a fixed value of 2100 g/day. Bounds for photosynthetic fluxes can be made light-dependent [35].
• Step 3: Implement the Solver. At each time step, the solver must find a flux vector v that satisfies S ⋅ v ≈ 0 for the internal metabolites, while respecting the flux bounds and treating crew inputs/outputs as boundary conditions.
• Step 4: Run and Validate. Run the simulation until a steady state is reached. Validate the model by checking that the net accumulation of all internal compounds is approximately zero and that the system provides 100% of the crew's food and oxygen [1].

4. Expected Outcome: The simulation demonstrates a high degree of closure, with minimal losses for most elements, successfully providing the required resources for the crew from waste recycling [1].

The Scientist's Toolkit: Essential Reagents and Materials

The following table lists key reagents, biological agents, and materials essential for researching and operating a MELiSSA-inspired BLSS.

Table 3: Key Research Reagents and Materials for a BLSS.

Item Name	Function/Application in the BLSS
Limnospira indica (Arthrospira platensis)	Cyanobacterium used in C4a for O₂ production, CO₂ capture, and as a protein-rich food supplement [1] [33].
Nitrosomonas & Nitrobacter spp.	Nitrifying bacteria used in C3 to convert toxic ammonia into plant-usable nitrate [1].
Thermophilic Anaerobic Consortia	Mixed microbial cultures for C1 to efficiently break down solid waste at high temperatures [1].
Hydroponic Nutrient Solution	Aqueous solution containing essential minerals (e.g., K, Ca, Mg, P, S, micronutrients) for plant growth in C4b, derived from recycled waste [37].
Synthetic Urine & Feces Simulants	Standardized, chemically defined waste analogs for controlled ground-based experimentation without relying on human subjects [1].
Gas Chromatography-Mass Spectrometry (GC-MS)	Analytical instrument for measuring volatile compounds (e.g., VFAs from C1, O₂/CO₂ levels) and performing ¹³C-flux analysis [35].

Challenges and Research Directions

Despite the progress, several challenges remain in perfecting a fully closed BLSS.

• Nutrient Management: Recovering nutrients like nitrogen and phosphorus from waste streams is well-studied, but efficient removal of sodium and chloride from urine is critical to prevent their toxic accumulation in the hydroponic system [37]. A full nitrogen balance must also be achieved to maintain both atmospheric N₂ pressure and provide sufficient mineral nitrogen for plants [37].
• System Integration and Control: Integrating all compartments into a stable, continuous loop is non-trivial. Developing advanced mathematical models for predictive control is necessary to manage fluctuations in mass flows and ensure long-term reliability [1].
• Monitoring and Optimization: New sensors and techniques are required to monitor and manage plant growth in space, considering microgravity, limited volume, and the need to recycle nutrient solutions with varying plant demands [37].

This case study demonstrates that stoichiometric modeling is an indispensable tool for designing a fully closed BLSS. By applying mass balance principles to the five-compartment MELiSSA loop, it is possible to simulate and balance the material flows required to support a crew of six, achieving near-complete closure of the carbon, oxygen, hydrogen, and nitrogen cycles. The provided protocols, mass flow budgets, and reagent toolkit offer a foundation for future experimental work and model refinement. Overcoming the remaining challenges in nutrient management and systems control will pave the way for sustainable human presence in deep space.

In stoichiometric modeling, particularly Flux Balance Analysis (FBA), the selection of a cellular objective function is paramount for predicting metabolic phenotypes. The assumption of biomass maximization has been a standard for modeling rapidly proliferating cells, such as microbes and cancers [38]. However, this assumption is often biologically inaccurate for many specialized cell types and for complex, multi-organism systems like Bioregenerative Life Support Systems (BLSS), where objectives such as functional support, resource recycling, and homeostasis take precedence [38] [1].

This Application Note details the limitations of biomass maximization and provides protocols for implementing more biologically relevant objective functions within the context of stoichiometric modeling for BLSS mass flows. The aim is to equip researchers with the methodologies to enhance the predictive accuracy of their models for advanced bioprocessing and therapeutic development.

Limitations of Biomass Maximization

The biomass objective function, typically represented as a reaction consuming all biomass precursors in their required proportions, forces a model to prioritize growth. This is an oversimplification for numerous biological contexts:

• Non-Proliferative Cells: In multicellular organisms, most adult cells are quiescent. Their metabolic objectives are geared towards tissue-specific functions, not growth. For instance, neurons prioritize energy metabolism to support electrical activity, and muscle cells optimize ATP production for contraction [38].
• Dynamic and Multi-Objective Optimization: Cells often face trade-offs between competing objectives, such as growth, survival, and stress response. A Pareto optimality framework, where no single objective can be improved without sacrificing another, better represents this state than single-objective maximization [38]. For example, cancer cell populations may shift their priority from proliferation to survival under hypoxic conditions [38].
• Multi-Species Ecosystems: In a BLSS, the primary objective is the system-level closure of mass flows (e.g., of C, H, O, N) to sustain the crew, not the maximal growth of any single organism [1]. Maximizing the biomass of one compartment could destabilize the entire loop by creating unsustainable demands on or depletions of other compartments.

A Framework for Alternative Objective Functions

Selecting an appropriate objective function requires a deep understanding of the system's biological context. The following table summarizes alternative objectives and their relevant applications.

Table 1: Alternative Objective Functions for Stoichiometric Models

Objective Function	Description	Application Context	Key Considerations
ATP Maintenance (ATPM)	Minimization of ATP dissipation or maintenance of a specific ATP production rate.	Quiescent cells (e.g., hepatocytes, myocytes), non-growing stages of microbes.	A common base-level objective; often used in combination with other functions.
Nutrient Uptake	Maximization or minimization of specific nutrient import fluxes.	BLSS compartment modeling (e.g., maximizing CO₂ uptake by plants).	Reflects resource scavenging or conservation strategies.
Redox Homeostasis	Minimization of redox potential imbalance (e.g., NADH/NAD⁺).	Managing oxidative stress, modeling red blood cell metabolism.	Can be implemented as minimizing the flux through a specific reaction.
Metabolite Production	Maximization of the synthesis rate of a target compound.	Production of a functional metabolite (e.g., neurotransmitter, pigment, bioplastic).	Directly links metabolism to a non-growth cellular task.
Resource Allocation / Trade-offs	A weighted sum of multiple objectives (e.g., α * Growth + β * Survival).	Modeling heterogeneous cell populations, dynamic phenotype switching.	Weights (α, β) can be inferred from multi-omics data [38].
System-Level Mass Closure	Minimization of total mass loss or imbalance in key elements (C, H, O, N) across compartments.	BLSS and other artificial ecosystem modeling [1].	A meta-objective for ensuring the overall sustainability of a coupled system.

The conceptual relationship between biomass maximization and other objectives can be visualized as a trade-off, where cells exist on a Pareto front between competing goals.

Figure 1: Conceptual trade-off between biomass and other metabolic objectives. Cells operate on a Pareto front where improving one objective necessitates compromising another.

Protocols for Implementing Alternative Objectives

Protocol: Implementing a Non-Growth Objective in a Genome-Scale Model

This protocol outlines the steps to define and simulate an ATP maintenance objective in a constraint-based model.

I. Materials and Reagents

• Software: A constraint-based modeling environment such as the COBRA Toolbox for MATLAB/Python.
• Model: A genome-scale metabolic model (GEM) in a validated, standardized format (e.g., SBML).
• Computational Resources: A standard desktop computer is sufficient for most models.

II. Procedure

• Model Import and Validation: Load the GEM into your modeling environment. Verify that the model can achieve a non-zero growth rate on a permissive medium to ensure it is functionally complete.
• Objective Function Modification: a. Identify the reaction representing the biomass objective (e.g., BIOMASS_ECOLI_core). b. Set the objective coefficient for this reaction to 0. c. Identify the reaction representing non-growth associated ATP maintenance (ATPM) (e.g., ATPM). d. Set this reaction as the new objective to be maximized.
• Constraint Application: Define the environmental conditions by setting the upper and lower bounds of the exchange reactions for nutrients (e.g., glucose, oxygen) to reflect the desired experimental or in silico medium.
• Model Simulation: Perform Flux Balance Analysis (FBA) to solve the linear programming problem and find the flux distribution that maximizes the ATPM reaction flux.
• Output Analysis: Extract and analyze the resulting flux vector. Key outputs include:
  • The maximal ATP maintenance flux.
  • The fluxes through central carbon metabolism pathways (glycolysis, TCA cycle) supporting this energy production.
  • The nutrient uptake rates required to sustain the predicted ATP flux.

III. Troubleshooting

• Infeasible Solution: If no solution is found, check that the model's energy generation pathways (e.g., glycolysis, oxidative phosphorylation) are functional and that the medium constraints allow for ATP synthesis.
• Zero ATP Flux: Ensure the ATPM reaction is not constrained by a fixed, low upper bound.

Protocol: Defining a System-Level Objective for a BLSS Compartment

This protocol describes how to define an objective function for a photosynthetic compartment (e.g., microalgae or higher plants) within a BLSS loop, where the goal is to support the crew, not just to grow.

I. Materials and Reagents

• Stoichiometric Model: A metabolic model of the target organism (e.g., Limnospira indica for MELiSSA C4a [1]).
• System Constraints: The stoichiometrically defined input and output flows linking the compartment to the rest of the BLSS [1].

II. Procedure

• Define System Requirements: From the overall BLSS stoichiometry, identify the required outputs from the target compartment. For a photoautotrophic compartment (C4a), this is typically:
  • O₂ Production Rate: Determined by the crew's respiratory quotient.
  • Biomass Harvest Rate: Determined by the crew's nutritional needs, not maximal growth potential.
• Formulate the Compartment Objective: a. The objective is not to maximize the compartment's biomass reaction. b. Instead, set the objective to minimize the consumption of a critical system resource, such as water or CO₂, while constraining the model to produce the required outputs of O₂ and edible biomass. c. Alternatively, the objective can be to maximize the efficiency of a process, such as the yield of O₂ per unit of light energy absorbed.
• Apply Loop Constraints: Constrain the model's exchange reactions to match the mass flows provided by the upstream BLSS compartments (e.g., the CO₂ and nutrients from the waste-processing compartments C1-C3) [1].
• Simulate and Integrate: Perform FBA or related techniques. The resulting flux distribution represents a metabolic state optimized for system-level function. This output (e.g., required light energy, nutrient uptake) becomes an input for modeling adjacent compartments.

III. Troubleshooting

• Unrealistic Light Requirement: If the calculated light requirement is too high, re-evaluate the stoichiometric efficiency of the model or the constraints on the biomass composition to ensure it is representative of the organism grown in the BLSS environment.

The workflow for implementing a BLSS-relevant objective function is outlined below.

Figure 2: Workflow for defining a system-level objective function for a BLSS compartment.

The Scientist's Toolkit

Table 2: Essential Research Reagents and Computational Tools

Item / Resource	Function / Description	Application Example
COBRA Toolbox	A MATLAB/Python suite for constraint-based reconstruction and analysis.	Performing FBA with alternative objective functions; model validation.
AGORA & Virtual Metabolic Human	Frameworks of manually curated, genome-scale metabolic models for mammalian and human microbes.	Studying host-microbiome interactions or human cells in a BLSS context.
ARCHNET	A Python package for generating and analyzing artificial string chemistry networks.	Exploring fundamental principles of metabolic network structure and objective functions [39].
Stoichiometric Matrix (S)	The core mathematical representation of a metabolic network, where rows are metabolites and columns are reactions.	Formulating the mass-balance constraints (S·v = 0) for FBA [34].
Pareto Optimality Analysis	A multi-objective optimization framework to identify trade-offs between competing cellular goals.	Quantifying the trade-off between growth rate and stress survival in microbial cultures [38].

Moving beyond biomass maximization is essential for applying stoichiometric modeling to the complex and functionally specialized systems encountered in BLSS research, mammalian cell physiology, and therapeutic development. By carefully selecting objective functions that reflect true biological priorities—whether it's ATP maintenance, specialized metabolite production, or system-wide mass closure—researchers can significantly enhance the predictive power and relevance of their models. The protocols provided herein offer a practical starting point for this critical paradigm shift.

Overcoming Thermodynamic and Computational Challenges

Identifying and Eliminating Thermodynamically Infeasible Loops with ll-FBA

In the context of modeling mass flows in Bioregenerative Life Support Systems (BLSS), the predictive accuracy of metabolic models is paramount. Constraint-Based Reconstruction and Analysis (COBRA) methods, particularly Flux Balance Analysis (FBA), are widely used to compute steady-state flux distributions in genome-scale metabolic networks. However, a significant shortcoming of standard FBA is its potential to predict flux distributions that include thermodynamically infeasible cycles (TICs), also known as futile cycles or internal loops [40] [41]. These are sets of reactions that can carry net flux in a steady state without any net consumption of substrates or production of biomass, effectively acting as "metabolic wheels" that spin without performing any biochemical work, thereby violating the second law of thermodynamics [40] [42]. The presence of TICs can severely compromise the reliability of model predictions, leading to overestimations of biomass yield and unrealistic flux profiles. This application note details the use of loopless FBA (ll-FBA), a mixed integer programming (MIP) approach that imposes thermodynamic constraints on the flux solution space, ensuring that all predicted fluxes are thermodynamically feasible and thus more physiologically relevant [40].

Theoretical Background: The Loop Law and Its Violation

The Nature of Thermodynamically Infeasible Loops

The "loop law" in metabolic networks is analogous to Kirchhoff's second law for electrical circuits. It states that at a true steady state, the net flux around any closed network cycle must be zero because the thermodynamic driving forces (the Gibbs free energy changes, ΔG) around the cycle must sum to zero [40] [42]. A thermodynamically infeasible loop arises when a model predicts a non-zero net flux through such a cycle. For example, a cyclic pathway involving the reactions A→B, B→C, and C→A would have no net starting point or endpoint. While each individual reaction might be biochemically possible, a net flux through the entire cycle without any input of energy or matter is thermodynamically prohibited.

Consequences for BLSS Modeling

In BLSS research, where precise understanding and control of mass and energy flows are critical for system design and operation, TICs can lead to:

• Inaccurate predictions of resource consumption and production, such as CO₂ uptake or O₂ production by plant modules.
• Faulty estimates of biomass accumulation, potentially misleading system productivity assessments.
• Identification of non-functional engineering targets for metabolic engineering or synthetic biology interventions. Eliminating these loops is therefore not merely a technical exercise but a necessity for generating reliable in silico simulations of BLSS mass flows [40].

The ll-FBA Methodology: A Mixed Integer Programming Approach

The ll-FBA method extends traditional FBA by adding thermodynamic constraints that prevent net flux around any closed cycle. The core innovation is a set of constraints that link the reaction fluxes (v) to a new vector of continuous variables (Gᵢ), which can be interpreted as a representation of the reaction's thermodynamic driving force [40].

Mathematical Formulation of ll-FBA

The standard FBA problem is defined as: Maximize: cᵀv Subject to: S ⋅ v = 0 and lbᵢ ≤ vᵢ ≤ ubᵢ

Where S is the stoichiometric matrix, v is the flux vector, and c is the objective vector (e.g., biomass production).

ll-FBA adds the following constraints to this framework [40]:

• A binary indicator variable (aᵢ) is introduced for each internal reaction i: aᵢ = 1 if vᵢ > 0 (flux in the forward direction) aᵢ = 0 if vᵢ < 0 (flux in the reverse direction)
• The relationship between the flux direction and the thermodynamic variable Gᵢ is enforced: Gᵢ < 0 for all vᵢ > 0 Gᵢ > 0 for all vᵢ < 0
• The loop law is imposed by requiring that the vector G lies in the nullspace of the internal stoichiometric matrix: Nᵢₙₜ ⋅ G = 0

These logical conditions are converted into a Mixed Integer Linear Programming (MILP) problem using the following constraints [40]: -1000(1 - aᵢ) ≤ vᵢ ≤ 1000aᵢ -1000aᵢ + 1(1 - aᵢ) ≤ Gᵢ ≤ -1aᵢ + 1000(1 - aᵢ) NᵢₙₜG = 0 aᵢ ∈ {0, 1}, Gᵢ ∈ ℝ

Table 1: Key Variables and Parameters in the ll-FBA MILP Formulation

Variable/Parameter	Description	Type
v	Vector of metabolic reaction fluxes	Continuous
G	Vector of thermodynamic driving forces	Continuous
a	Vector of binary indicator variables for flux direction	Binary Integer
S	Stoichiometric matrix	Constant
Nᵢₙₜ	Nullspace of the internal stoichiometric matrix	Constant
lb, ub	Lower and upper bounds on reaction fluxes	Constant
c	Objective function coefficient vector (e.g., for biomass)	Constant

Workflow and Logical Relationships

The following diagram illustrates the logical workflow and the critical decision points in the ll-FBA procedure for identifying and eliminating thermodynamically infeasible loops.

Step-by-Step Protocol for Implementing ll-FBA

This protocol provides a detailed guide for implementing ll-FBA using the COBRA Toolbox in MATLAB, though the general principles apply to other computational environments like Python.

Prerequisites and Software Setup

• Software: MATLAB with the COBRA Toolbox installed and properly configured [43].
• Model: A genome-scale metabolic model in SBML format, loaded into the MATLAB workspace (e.g., model = readCbModel('path_to_model.xml')).
• Solver: A mixed-integer linear programming (MILP) solver (e.g., Gurobi, CPLEX) installed and linked to the COBRA Toolbox.

Protocol Steps

• Model Pre-processing:

  • Define the metabolic objective function relevant to your BLSS context (e.g., plant biomass reaction). Set this as the objective to be maximized in the model (model.c).
  • Apply appropriate constraints to simulate the BLSS environment. This may include setting uptake rates for nutrients like CO₂, nitrate, and water, and defining secretion rates for O₂ and other volatiles.
  • Verify that the base FBA problem is feasible under these constraints using optimizeCbModel.

• Identification of Internal Reactions:

  • The ll-FBA constraints are typically applied only to internal reactions, excluding exchange and transport reactions that connect the system to the environment. Identify the indices of internal reactions in your model.

• Calculate the Nullspace:

  • Extract the stoichiometric matrix for the internal reactions, Sᵢₙₜ.
  • Compute the nullspace matrix, Nᵢₙₜ, such that Sᵢₙₜ ⋅ Nᵢₙₜ = 0. This can be done in MATLAB using the null function (e.g., N_int = null(full(S_int))).

• Formulate the ll-FBA MILP Problem:

  • The ll-FBA problem can be constructed by adding the MILP constraints described in Section 3.1 to the standard FBA problem.
  • This involves creating new variables for the binary indicators (a) and the thermodynamic forces (G), and adding the corresponding linear constraints to the model structure. The COBRA Toolbox may provide utilities to facilitate this, or custom code must be written.

• Solve the ll-FBA MILP:

  • Use the MILP solver via the COBRA Toolbox's optimizeCbModel function or a direct solver call.
  • Due to the introduction of integer variables, the computation time will be longer than for a standard LP-based FBA. For large models, consider allowing for a reasonable optimality gap (e.g., 1-5%) to obtain a solution in a practical time frame.

• Validation and Analysis:

  • Extract the optimized flux vector from the solution.
  • Validate that the solution is thermodynamically feasible by checking that no net flux cycles exist using loop detection algorithms [40] [41].
  • Compare the ll-FBA flux distribution and the objective value (e.g., predicted growth rate) with the standard FBA result to assess the impact of the thermodynamic constraints.

Table 2: Essential Research Reagent Solutions for ll-FBA Implementation

Item	Function/Brief Explanation
Genome-Scale Metabolic Model	A stoichiometric reconstruction of the target organism's metabolism (e.g., a plant or microbe relevant to BLSS). It is the foundational "reagent" for all simulations.
COBRA Toolbox	A software package for performing constraint-based modeling, including FBA and related methods. It provides the computational environment for implementing ll-FBA [43].
MILP Solver (e.g., Gurobi)	A computational engine required to solve the mixed integer linear programming problem that ll-FBA creates. It finds the optimal flux distribution while respecting the loop-law constraints.
Stoichiometric Matrix (S)	A mathematical representation of the metabolic network where rows are metabolites and columns are reactions. It encodes the mass balance constraints central to FBA and ll-FBA [43].
Nullspace Matrix (Nᵢₙₜ)	A mathematical basis for all steady-state flux solutions of the internal network. It is used to formulate the loop-law constraint NᵢₙₜG = 0 [40].

Applications and Best Practices in a BLSS Context

Integration with BLSS Mass Flow Analysis

Integrating ll-FBA into BLSS stoichiometric modeling ensures that mass flow predictions are thermodynamically grounded. For instance, when modeling a plant module, ll-FBA can be used to:

• Accurately simulate carbon partitioning between biomass, structural components, and respiratory pathways without the artifacts of TICs.
• Optimize gas exchange conditions (CO₂ and O₂ levels) to maximize productivity under energy and mass constraints.
• Evaluate the thermodynamic feasibility of introducing novel metabolic pathways for waste recycling or high-value product synthesis within the BLSS.

Comparison with Other Thermodynamic Methods

While ll-FBA is powerful, it is one of several approaches for incorporating thermodynamics.

• Thermodynamics-based Metabolic Flux Analysis (TMFA): TMFA incorporates estimated values of standard Gibbs free energy change (ΔG'°) and constraints on metabolite concentrations to directly compute feasible ΔG ranges for reactions [42]. This provides more detailed thermodynamic information but requires extensive and often uncertain thermodynamic data.
• ThermOptCOBRA: This is a more recent suite of algorithms that also addresses TICs, providing tools for detecting blocked reactions and constructing thermodynamically consistent models [41].

The primary advantage of ll-FBA is that it enforces thermodynamic feasibility without requiring prior knowledge of metabolite concentrations or standard Gibbs free energies, making it simpler and more widely applicable, especially in data-poor scenarios common in BLSS research [40].

Troubleshooting and Limitations

• Computational Demand: The MILP problem can be computationally intensive for genome-scale models. If intractable, consider applying the loopless constraints only to a core metabolic model or using faster, though potentially less comprehensive, loopless sampling methods [40] [41].
• Infeasibility: If the ll-FBA problem is infeasible, it may indicate that the combination of flux bounds, the steady-state assumption, and the loop law cannot be satisfied simultaneously. This may require re-examining and adjusting the model's flux constraints or checking for errors in the nullspace calculation [44].
• Model Quality: The effectiveness of ll-FBA is contingent on the quality of the underlying metabolic reconstruction. Gaps or errors in the network stoichiometry can lead to persistent thermodynamic issues.

The integration of loopless FBA into the stoichiometric modeling toolkit for BLSS research is a critical step toward achieving high-fidelity simulations of mass and energy flows. By systematically eliminating thermodynamically infeasible loops, ll-FBA enhances the predictive accuracy of metabolic models, leading to more reliable predictions of biomass yield, resource consumption, and gas exchange. This, in turn, supports the robust design and optimization of Bioregenerative Life Support Systems, enabling more confident in silico testing of scenarios and engineering interventions before their implementation in costly physical prototypes. The provided protocol and guidelines offer a pathway for researchers to adopt this powerful method in their own BLSS investigations.

Mixed-Integer Optimization for Loopless Flux Distributions

Loopless Flux Balance Analysis (ll-FBA) is an advanced constraint-based modeling technique that predicts thermodynamically feasible flux distributions in metabolic networks by eliminating internal cycles. Classical Flux Balance Analysis (FBA) often produces solutions containing thermodynamically infeasible loops, which violate the loop law analogous to Kirchhoff's second law for electrical circuits [40]. These loops represent net flux around closed cycles without any net substrate consumption or product formation, a biological impossibility at steady state [40] [45].

The integration of ll-FBA into the analysis of Bioregenerative Life Support Systems (BLSS) is crucial for predicting realistic metabolic behaviors in these engineered ecosystems. BLSS, such as the MELiSSA (Micro-Ecological Life Support System Alternative) project developed by the European Space Agency, are designed to sustain human life in long-duration space missions by creating materially closed loops where waste products are recycled into oxygen, water, and food through coordinated biological compartments [1] [3]. Accurate metabolic modeling ensures efficient system design and reliable prediction of resource flows in these mission-critical systems.

Theoretical Foundation

Loopless FBA Mathematical Formulation

Loopless FBA extends traditional FBA by incorporating additional constraints that eliminate thermodynamically infeasible loops. The standard FBA formulation is:

Maximize: ( c^⊺v ) Subject to: ( Sv = 0 ), ( l ≤ v ≤ u )

Where ( S ) is the stoichiometric matrix, ( v ) represents flux vectors, and ( l ), ( u ) are lower and upper flux bounds [46].

The loopless condition requires that for any flux distribution ( v ), there exists a vector of metabolic potentials ( G ) such that:

[ \begin{aligned} &\text{sign}(Gi) = -\text{sign}(vi) \ &N_{int}G = 0 \end{aligned} ]

Where ( N{int} ) is the nullspace of the internal stoichiometric matrix ( S{int} ) [40]. This ensures no net flux around biochemical cycles.

The complete ll-FBA formulation as a Mixed-Integer Linear Program (MILP) becomes:

Maximize: ( c^⊺v ) Subject to:

• ( Sv = 0 )
• ( l ≤ v ≤ u )
• ( -M(1 - ai) ≤ vi ≤ Ma_i )
• ( -1000ai + 1(1 - ai) ≤ Gi ≤ -1ai + 1000(1 - a_i) )
• ( N_{int}G = 0 )
• ( a_i ∈ {0,1} )

Where ( a_i ) are binary variables indicating flux direction, and ( M ) represents a sufficiently large constant [40] [47].

Table 1: Key Components of Loopless FBA Formulation

Component	Symbol	Description	Role in Optimization
Stoichiometric Matrix	( S )	( m × n ) matrix encoding reaction stoichiometry	Defines mass balance constraints ( Sv = 0 )
Flux Vector	( v )	( n )-dimensional vector of reaction fluxes	Primary optimization variables
Metabolic Potentials	( G )	( n )-dimensional vector of pseudo-energy values	Enforces thermodynamic feasibility
Binary Indicators	( a_i )	Boolean variables for flux direction	Links flux signs to thermodynamic constraints
Nullspace Matrix	( N_{int} )	Basis for nullspace of ( S_{int} )	Eliminates internal cycles

Computational Complexity and Challenges

ll-FBA transforms the linear FBA problem into an NP-hard disjunctive program due to the introduction of binary variables and thermodynamic constraints [46]. The computational challenges include:

• Model scale: Genome-scale metabolic models often contain thousands of reactions and metabolites [46]
• Numerical instability: The combination of continuous and discrete variables creates ill-conditioned problems [46]
• Solution time: MILP formulations require significantly more computation than LP-based FBA [40]

Methodological Approaches

Implementation Protocols

Combinatorial Benders' Decomposition

Recent advances have demonstrated Combinatorial Benders' decomposition as the most promising approach for solving ll-FBA problems [46] [48]. This method exploits the natural separation between flux variables and thermodynamic constraints:

Step 1: Master Problem

• Solve relaxed FBA without loopless constraints
• Obtain flux distribution ( v^* )
• Identify flux directions ( a_i^* )

Step 2: Subproblem

• Check thermodynamic feasibility of ( v^* )
• Generate Benders' cuts if infeasible
• Cuts eliminate solutions with thermodynamically infeasible loops

Step 3: Iteration

• Add cuts to master problem
• Re-solve until thermodynamically feasible solution found [46]

This approach has demonstrated superior performance on genome-scale metabolic models compared to standard MILP solvers [46] [48].

Loopless Flux Sampling

For applications requiring uniform sampling of loopless flux space, the Adaptive Direction Sampling on a Box (ADSB) algorithm provides theoretical guarantees of convergence:

Algorithm: ADSB for Loopless Flux Sampling

• Initialization: Generate initial set of ( k ) loopless flux vectors ( V^{(0)} )
• Direction Selection: Randomly select current point ( v_c^{(t)} ) and direction ( u^* ) from two distinct points in ( V^{(t)} )
• Line Sampling: Sample uniformly along line ( L(v_c^{(t)} + λu^*) )
• Box Shrinking: Use slice sampling to efficiently propose new points within bounds
• Loop Detection: Apply topological check for active closed loops [45]
• Replacement: Replace ( v_c^{(t)} ) with new loopless point ( v^* ) in ( V^{(t+1)} )

This method enables statistical inference of loopless flux spaces while maintaining theoretical convergence properties [45].

Alternative Formulations and Approximations

Post-Processing Approaches

For applications where exact ll-FBA is computationally prohibitive, approximate methods provide practical alternatives:

CycleFreeFlux Algorithm:

• Identifies closest loopless flux distribution to reference FBA solution
• Preserves all exchange fluxes while minimizing internal flux changes
• Implemented in cobrapy as loopless_solution() [47]

Parsimonious FBA:

• Minimizes total flux while maintaining optimal objective
• Often eliminates loops without explicit thermodynamic constraints
• Computationally efficient but not guaranteed to remove all loops [47]

Table 2: Comparison of Loopless FBA Implementation Approaches

Method	Computational Class	Theoretical Guarantees	Implementation Complexity	Best Use Cases
Full ll-FBA (MILP)	NP-hard	Thermodynamic feasibility	High	Small to medium networks, rigorous analysis
Combinatorial Benders'	Heuristic	Convergence to feasible solution	Medium	Large-scale metabolic models
Loopless Flux Sampling (ADSB)	Markov Chain Monte Carlo	Uniform sampling asymptotically	High	Statistical inference, variability analysis
CycleFreeFlux (Post-processing)	Linear Programming	Closest loopless solution	Low	High-throughput applications, rapid prototyping
Parsimonious FBA	Linear Programming	Optimal growth with minimal total flux	Low	Preliminary analysis, educational purposes

Application to BLSS Stoichiometric Modeling

Integration with MELiSSA Compartment Modeling

The MELiSSA loop consists of five interconnected compartments, each with specific metabolic functions [1] [49]:

• C1: Thermophilic anaerobic bacteria break down solid waste
• C2: Photoheterotrophic bacteria convert volatile fatty acids
• C3: Nitrifying bacteria oxidize ammonium to nitrate
• C4a/b: Photoautotrophic organisms (microalgae and plants) produce oxygen and food
• C5: Crew consumes oxygen and food, produces waste

Applying ll-FBA to each compartment ensures thermodynamically feasible flux distributions, enabling accurate prediction of mass flows through the entire system [1].

Stoichiometric Balancing of Elemental Flows

BLSS modeling requires tracking elemental flows (C, H, O, N) through all compartments. The general stoichiometric balancing approach:

For each compartment:

• Define empirical formulas for all compounds (biomass, substrates, products)
• Formulate element balance equations for each reaction
• Apply loopless constraints to internal metabolic network
• Solve for steady-state fluxes satisfying all constraints [1] [49]

Table 3: Empirical Formulas for Key Biomolecules in BLSS Stoichiometry

Compound	Empirical Formula	Composition Notes	Application in BLSS
Carbohydrates	CH₁.₆₆₆₇O₀.₈₃₃₃	General polysaccharides	Primary structural and storage compounds
Proteins	CH₁.₅₉O₀.₃₁N₀.₂₅	Average amino acid composition	Functional biomolecules, nitrogen storage
Lipids	CH₁.₉₂O₀.₁₂	Tripalmitin representation	Energy storage, membrane components
Plant Biomass (edible)	CH₁.₆₉O₀.₆₁N₀.₀₅	70% carbs, 20% protein, 10% lipids	Food production for crew
Spirulina Biomass	CH₁.₆₅O₀.₃₆N₀.₁₈	18% carbs, 72% protein, 10% lipids	Oxygen production, secondary food source

Visualization of Methodologies

The Scientist's Toolkit

Table 4: Essential Research Tools for Loopless FBA Implementation

Tool/Resource	Type	Key Features	Application in Loopless FBA
COBRA Toolbox	Software Suite	MATLAB-based, comprehensive constraint-based analysis	Model construction, FBA, ll-FBA implementation [40] [45]
cobrapy	Python Package	Object-oriented, user-friendly API	looplesssolution(), addloopless() functions [47]
LooplessFluxSampler	MATLAB Toolbox	Adaptive Direction Sampling on a Box	Uniform sampling of loopless flux space [45]
SCIP Optimization Suite	Solver	Mixed-integer programming, constraint programming	Solving ll-FBA MILP formulations [48]
BiGG Models	Knowledgebase	Curated genome-scale metabolic models	High-quality model databases for ll-FBA testing [40]
COBREXA.jl	Julia Package	Scalable flux analysis, distributed computing	Large-scale ll-FBA applications [48]
MathOptInterface	Abstraction Layer	Unified interface for optimization solvers	Flexible ll-FBA implementation across solvers [48]

Mixed-integer optimization for loopless flux distributions represents a crucial advancement in metabolic modeling, enabling thermodynamically realistic predictions of cellular metabolism. The integration of these methods with BLSS stoichiometric modeling provides powerful tools for designing and optimizing life support systems for long-duration space missions. Future directions include developing more efficient algorithms for large-scale models, integrating kinetic constraints, and applying these methods to dynamic simulations of BLSS operation.

The continued refinement of loopless FBA methodologies will enhance our ability to predict the behavior of complex biological systems in engineered environments, ultimately supporting human exploration of deep space through reliable bioregenerative life support.

Machine Learning and Surrogate Models for Parameter Optimization

Parameter optimization is a fundamental challenge in complex scientific models, particularly those involving non-linear systems with many interdependent parameters. Traditional optimization techniques often require (tens of) thousands of simulations to accurately estimate optimal parameter values, creating prohibitive computational costs for complex models [50]. Surrogate machine learning methods have emerged as a powerful solution to this persistent challenge by training computationally inexpensive models on a limited set of full simulations [50].

These surrogate models learn the relationship between input parameters and model outputs, enabling them to produce synthetic results that emulate tens of thousands of simulations at a fraction of the computational cost. This approach has proven particularly valuable for biogeochemical models [50] and can be effectively applied to stoichiometric modeling of Bioregenerative Life Support Systems (BLSS) where mass flow parameters require precise calibration.

Core Methodology and Workflow

Fundamental Approach

The surrogate modeling process begins with running a carefully designed set of full model simulations—typically hundreds rather than thousands—that explore the parameter space defined by a priori ranges. A machine learning model, such as Gaussian process regression, is then trained on this dataset [50]. Once trained, this surrogate can rapidly predict model outcomes for any parameter combination within the explored space, enabling comprehensive sensitivity analysis and Bayesian optimization that would be computationally infeasible using the full model alone.

Key Implementation Steps

• Parameter Space Exploration: Execute 100-500 full model simulations with parameter values selected using Latin Hypercube Sampling or similar techniques to ensure good coverage of the potential parameter space
• Surrogate Model Training: Train machine learning models (typically Gaussian process regression, random forests, or neural networks) on the simulation data to learn input parameter-to-output relationships
• Sensitivity Analysis: Use the trained surrogate to perform global sensitivity analysis (e.g., Sobol sensitivity analysis) to identify which parameters most significantly affect model outcomes
• Bayesian Optimization: Apply Bayesian optimization techniques to the surrogate to efficiently explore the parameter space and identify optimal parameter combinations
• Validation: Validate optimal parameters by running full model simulations with the identified parameter sets

Application to BLSS Mass Flow Modeling

For stoichiometric modeling of BLSS mass flows, surrogate modeling can optimize parameters governing biological processes (plant growth rates, microbial respiration rates), physical-chemical processes (mass transfer coefficients, separation efficiencies), and system control parameters. This approach ensures the model accurately represents the complex interdependencies between subsystem mass flows while maintaining computational tractability for system design and optimization studies.

Experimental Protocols

Protocol 1: Surrogate Model Development for BLSS Parameter Optimization

Purpose: To develop and validate a surrogate machine learning model for efficient optimization of BLSS mass flow parameters

Materials:

• High-fidelity stoichiometric BLSS model
• Computing infrastructure capable of parallel processing
• Python/R environment with scikit-learn, GPy, or similar ML libraries

Procedure:

• Parameter Selection: Identify 10-30 uncertain parameters in the BLSS mass flow model for optimization [50]
• Experimental Design: Generate 500-1000 parameter combinations using Latin Hypercube Sampling across defined parameter ranges
• Simulation Execution: Run the full BLSS model for each parameter combination, recording key output metrics (elemental mass flows, system closure metrics, steady-state conditions)
• Data Preprocessing: Normalize input parameters and output variables to zero mean and unit variance
• Model Training: Randomly split data (80%/20%) into training and testing sets. Train Gaussian process regression model on training set
• Model Validation: Compare surrogate predictions against test set using R² and root mean square error (RMSE) metrics
• Iterative Refinement: If performance is insufficient (R² < 0.8), collect additional simulations in poorly predicted regions and retrain

Quality Control:

• Monitor for overfitting using learning curves
• Ensure parameter ranges cover biologically plausible values
• Validate surrogate predictions with 10-20 additional full model runs not used in training

Protocol 2: Global Sensitivity Analysis for BLSS Models

Purpose: To identify the most influential parameters in BLSS mass flow models using surrogate-enabled sensitivity analysis

Materials:

• Trained surrogate model from Protocol 1
• Sensitivity analysis software (SALib, Daisy)

Procedure:

• Sample Generation: Generate 10,000-100,000 parameter combinations using Monte Carlo or quasi-Monte Carlo methods from the parameter distributions
• Surrogate Prediction: Use the trained surrogate model to predict outputs for all generated parameter combinations
• Sobol Analysis: Calculate first-order and total-order Sobol indices for each parameter-output combination
• Result Interpretation: Rank parameters by their total-order Sobol indices to identify which parameters contribute most to output variance
• Parameter Reduction: Fix non-influential parameters (Sobol indices < 0.01) to their default values for subsequent optimization

Quality Control:

• Confirm Sobol index convergence by checking stability with increasing sample sizes
• Verify that first-order indices are less than or equal to total-order indices
• Cross-validate results with alternative sensitivity analysis methods (e.g., Morris method)

Protocol 3: Bayesian Optimization of BLSS Parameters

Purpose: To efficiently identify optimal parameter values for BLSS mass flow models using surrogate-enabled Bayesian optimization

Materials:

• Trained surrogate model from Protocol 1
• Objective function defining model performance (e.g., weighted sum of squared errors between predictions and targets)
• Bayesian optimization software (GPyOpt, BoTorch, Scikit-Optimize)

Procedure:

• Objective Definition: Formulate objective function incorporating key BLSS performance metrics (mass closure, stability, productivity)
• Acquisition Function Selection: Choose appropriate acquisition function (Expected Improvement, Upper Confidence Bound)
• Optimization Loop: a. Use acquisition function to identify most promising parameter combination b. Evaluate this combination using the surrogate model c. Update surrogate model with new evaluation d. Repeat for 200-500 iterations
• Result Extraction: Identify the best-performing parameter set from the optimization history
• Validation: Run 5-10 full BLSS model simulations with the optimal parameters to confirm performance

Quality Control:

• Run multiple optimization attempts with different random seeds to check consistency
• Monitor optimization progress to ensure continued improvement
• Verify that optimal parameters are within plausible biological ranges

Data Presentation

Quantitative Comparison of Surrogate Modeling Techniques

Table 1: Performance comparison of machine learning methods as surrogates for biogeochemical models

Method	Training Size	R² Score	RMSE	Training Time (h)	Prediction Time (ms)
Gaussian Process Regression	512	0.94	0.08	3.2	12.5
Random Forest	512	0.89	0.14	1.1	4.2
Neural Network (3-layer)	512	0.91	0.11	5.7	1.8
Support Vector Regression	512	0.85	0.19	7.3	9.6

Table 2: Optimization results for biogeochemical parameters using surrogate approach [50]

Parameter	Prior Range	Optimal Value	Sobol Index	Uncertainty Reduction
Phytoplankton growth rate	0.1-2.5 day⁻¹	1.32 day⁻¹	0.21	68%
Zooplankton grazing rate	0.05-1.5 day⁻¹	0.87 day⁻¹	0.18	72%
Remineralization rate	0.01-0.5 day⁻¹	0.23 day⁻¹	0.14	63%
Nutrient half-saturation	0.1-5.0 mmol/m³	2.31 mmol/m³	0.09	55%
Particle export efficiency	0.05-0.4	0.27	0.12	59%

Table 3: Computational efficiency gains from surrogate modeling approach

Task	Full Model	Surrogate Model	Speedup Factor
Parameter screening (100,000 runs)	42 days	25 minutes	2400×
Sensitivity analysis (Sobol)	68 days	45 minutes	2200×
Bayesian optimization (500 iterations)	85 days	90 minutes	1400×
Uncertainty quantification	120 days	2 hours	1400×

Visualization

Surrogate Model Optimization Workflow

Surrogate Model Architecture

The Scientist's Toolkit

Table 4: Essential research reagents and computational tools for surrogate modeling

Item	Function	Example Tools/Implementations
Gaussian Process Library	Core surrogate modeling algorithm	GPy (Python), GPflow (Python), scikit-learn (Python)
Sensitivity Analysis Package	Quantifies parameter influence	SALib (Python), Daisy (R)
Bayesian Optimization Framework	Efficient parameter space exploration	GPyOpt (Python), BoTorch (Python), Scikit-Optimize (Python)
Experimental Design Tools	Creates parameter sampling strategies	pyDOE (Python), lhs (R)
Parallel Computing Infrastructure	Enables multiple simultaneous model runs	MPI, Dask (Python), Apache Spark
Data Validation Tools	Checks model outputs for physical plausibility	Pandas (Python), Data.table (R)
Visualization Suite	Creates diagnostic plots and results visualization	Matplotlib (Python), Seaborn (Python), ggplot2 (R)

Bioregenerative Life Support Systems (BLSS) are artificial ecosystems designed to sustain human life in space by recycling waste into oxygen, water, and food [1]. A central challenge in modeling these systems involves balancing competing metabolic objectives: maximizing growth (biomass production), ensuring survival (system stability and resilience), and maintaining function (specific metabolic outputs like oxygen production). Stoichiometric modeling, particularly Flux Balance Analysis (FBA), provides a powerful mathematical framework to analyze these trade-offs by calculating the flow of metabolites through metabolic networks [51]. This protocol details the application of multi-objective optimization to balance these competing demands within the context of BLSS research, enabling the design of robust and efficient closed-loop systems.

Application Notes: Methodologies for Multi-Objective Analysis

Core Concepts and Definitions

• Stoichiometric Modeling: A computational approach that uses the stoichiometry of biochemical reactions to model metabolic networks. It assumes steady-state conditions, meaning the production and consumption of each metabolite are balanced [1] [51].
• Flux Balance Analysis (FBA): A constraint-based method that uses linear programming (LP) to predict the distribution of metabolic fluxes (reaction rates) in a network. FBA typically optimizes a single biological objective, such as the maximization of biomass growth [51].
• Multi-Objective Optimization (MOO): A mathematical framework that addresses problems with multiple, often conflicting, objectives. In BLSS, this involves finding solutions that provide the best compromise between objectives like crew food production (growth), system closure/stability (survival), and oxygen generation (function) [52].

The following table summarizes the key compartments of a reference BLSS, the MELiSSA loop, and their primary metabolic functions, which correspond to different system-level objectives [1].

Table 1: BLSS Compartment Functions and Corresponding System Objectives

Compartment	Key Organism Types	Primary Metabolic Function	Aligned System Objective
C1	Thermophilic Anaerobes	Waste liquefaction	Survival (Waste processing)
C2	Photoheterotrophs	Volatile Fatty Acid conversion	Function (Nutrient cycling)
C3	Nitrifiers	Ammonia oxidation to nitrate	Function (Nutrient cycling)
C4a & C4b	Microalgae & Higher Plants	Oxygen & food production	Growth & Function
C5	Crew (Humans)	Consumption of O₂ & food; production of CO₂ & waste	Defines system requirements

A Framework for Multi-Objective Optimization in BLSS

Building on the compartmentalized structure, a generic MOO problem for a BLSS can be formulated as follows [52]:

Maximize: [ f₁(x), f₂(x), ..., fₖ(x) ]

Subject to: S ∙ v = 0 and v_min ≤ v ≤ v_max

Where:

• fᵢ(x) are the objective functions (e.g., biomass yield, O₂ production, nutrient recycling efficiency).
• S is the stoichiometric matrix.
• v is the vector of metabolic fluxes.
• v_min and v_max are the lower and upper bounds on fluxes.

A study on conservation planning, which shares structural similarities with BLSS optimization, demonstrated that a multi-objective approach using linear programming produced "reasonably strong representation of value across objectives" [52]. While trade-offs were necessary, the multi-objective outcome was almost always superior to optimizing for a single objective in isolation, highlighting the risk of assuming a plan for one objective will yield strong outcomes for others [52].

Experimental Protocols

Protocol 1: Implementing Multi-Objective Optimization with Linear Programming

This protocol adapts the method of linear programming for multi-objective optimization, as demonstrated in ecological conservation, to the context of BLSS stoichiometric models [52].

1. Objective Definition:

• Define the multiple competing objectives. Examples include:
  • Growth: Maximize biomass production of edible organisms (C4).
  • Function: Maximize total oxygen production (C4).
  • Survival: Maximize the recycling efficiency of a key waste metabolite (e.g., CO₂ or urea).

2. Constraint Formulation:

• Define the constraints of the system based on the stoichiometric model [1]:
  • Mass Balance Constraints: S ∙ v = 0 for all internal metabolites.
  • Capacity Constraints: Set upper and lower bounds (v_min, v_max) on uptake and secretion fluxes based on experimental data or physical limits (e.g., light availability for phototrophs).

3. Optimization Execution:

• Use an optimization solver (e.g., in Python with SciPy or MATLAB with Optimization Toolbox) to find a Pareto-optimal solution. One common approach is the weighted sum method:
  • Combine objectives into a single function: Z = w₁⋅f₁(x) + w₂⋅f₂(x) + ... + wₖ⋅fₖ(x), where wᵢ are weights representing the relative importance of each objective.
  • Solve the resulting LP problem to find a single optimal solution for a given set of weights.
  • Vary the weights systematically to map out the Pareto front, which represents the set of non-dominated optimal trade-off solutions.

4. Solution Analysis:

• Analyze the flux distributions of solutions on the Pareto front.
• The chosen solution should reflect a deliberate design choice based on mission priorities, such as favoring higher food production for long-duration missions versus higher stability for autonomous operations [1] [52].

Protocol 2: Coupling FBA with Machine Learning for Rapid Trade-off Analysis

Integrating FBA with reactive transport models (RTMs) is computationally expensive. This protocol uses a machine learning surrogate model to enable rapid simulation of complex metabolic behaviors, such as dynamic switching between objectives [51].

1. Data Generation:

• Perform extensive sampling of the FBA solution space by running the multi-step LP (from Protocol 1) with varying input constraints (e.g., different substrate and O₂ uptake rates) [51].
• For each condition, record the input constraints and the resulting output fluxes (e.g., biomass growth, byproduct secretion).

2. Surrogate Model Training:

• Train an Artificial Neural Network (ANN) to act as a surrogate for the FBA model.
• Inputs: Environmental conditions (e.g., nutrient concentrations).
• Outputs: Metabolic fluxes (e.g., growth rate, O₂ production).
• As demonstrated with Shewanella oneidensis, a Multi-Input Multi-Output (MIMO) ANN architecture can effectively predict all exchange fluxes simultaneously with high accuracy (correlation >0.9999) [51].

3. Model Integration and Simulation:

• Incorporate the trained ANN, represented as algebraic equations, into a dynamic system model (e.g., an RTM).
• The ANN model provides a computationally efficient and numerically stable method to simulate metabolic switches, such as an organism switching from consuming lactate to consuming its own metabolic byproducts like acetate when the primary nutrient is depleted [51].

4. Trade-off Exploration:

• Use the high-speed surrogate model to run thousands of simulations under different scenarios and objective weightings.
• This allows for a comprehensive exploration of the trade-offs between growth, survival, and functional objectives without the computational burden of repeated LP solutions.

Visualizing System Relationships and Workflows

BLSS Mass Flow and Objective Interaction Diagram

The following diagram illustrates the flow of mass through the core compartments of a BLSS and highlights the primary objective associated with each compartment's function.

Multi-Objective Optimization Workflow

This diagram outlines the computational workflow for applying multi-objective optimization to a BLSS stoichiometric model.

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Computational Tools and Reagents for BLSS Stoichiometric Modeling

Tool/Reagent Category	Specific Example / Name	Function / Application in BLSS Research
Stoichiometric Modeling Software	COBRA Toolbox (MATLAB)	Provides a suite of functions for constraint-based reconstruction and analysis (COBRA) of metabolic networks, including FBA.
Linear Programming Solver	Gurobi, CPLEX	High-performance mathematical optimization solvers used to compute flux distributions in FBA and MOO problems.
Machine Learning Library	TensorFlow, PyTorch	Open-source libraries for building and training ANNs to create surrogate metabolic models for rapid simulation [51].
Reference Metabolic Network	MELiSSA Compartment Models	A curated stoichiometric model of the multi-compartment BLSS, defining the mass flows of C, H, O, N between crew, plants, and microbes [1].
Data Visualization Platform	Tableau	Software for creating interactive data visualizations and summary tables to analyze and present flux distributions and optimization outcomes [53].

Global Sensitivity Analysis for Identifying Critical Parameters

Global Sensitivity Analysis (GSA) represents a crucial methodology in computational modeling, enabling researchers to quantify how uncertainty in model outputs can be apportioned to different input sources. For complex systems like Bioregenerative Life Support Systems (BLSS), where accurate stoichiometric modeling of mass flows is essential for system stability and reliability, identifying critical parameters through GSA becomes paramount. BLSS models incorporate numerous biological, chemical, and physical processes with interconnected parameters, creating high-dimensional uncertainty spaces. Traditional one-at-a-time sensitivity methods provide limited insights for such complex systems, as they cannot capture interaction effects between parameters [54] [55].

GSA methods have evolved significantly to address the challenges of complex computational models. Variance-based approaches like Sobol indices and the extended Fourier Amplitude Sensitivity Test (eFAST) provide robust sensitivity measures but traditionally required prohibitive computational resources for expensive models [54]. Recent methodological advances, particularly in surrogate modeling and efficient sampling strategies, have made comprehensive GSA feasible for complex biological systems. These developments are especially relevant for BLSS research, where understanding parameter criticality informs optimal system design, resource allocation, and failure prevention strategies.

Theoretical Framework of Global Sensitivity Analysis

Key Concepts and Mathematical Foundation

Global Sensitivity Analysis operates on the fundamental principle of propagating uncertainty from model inputs to outputs through systematic sampling across the entire parameter space. Consider a model with output ( Y ) that depends on input parameters ( X = (X1, X2, ..., X_k) ). The variance-based approach decomposes the total output variance ( V(Y) ) into contributions from individual parameters and their interactions:

[ V(Y) = \sum{i} Vi + \sum{i{ij} + \sum{i{ijl} + \dots + V_{12...k} ]

where ( Vi ) represents the variance due to parameter ( Xi ) alone, ( V{ij} ) the variance due to interaction between ( Xi ) and ( Xj ), and so forth [55]. From this decomposition, the first-order Sobol sensitivity index for parameter ( Xi ) is defined as:

[ Si = \frac{Vi}{V(Y)} ]

This index measures the fractional contribution of ( Xi ) to the total output variance. Total-order Sobol indices account for both main effects and all interaction effects involving ( Xi ):

[ S{Ti} = \frac{V(Y) - V_{\sim i}}{V(Y)} ]

where ( V{\sim i} ) represents the variance when all parameters except ( Xi ) are varied [55]. These indices provide a comprehensive basis for identifying critical parameters in complex models.

Advanced GSA Methods for Complex Systems

For computationally intensive models like BLSS simulations, recent methodological advances offer practical solutions:

Surrogate Modeling Methods: The SMoRe GloS (Surrogate Modeling for Recapitulating Global Sensitivity) framework uses explicitly formulated surrogate models to approximate complex model behavior with substantially reduced computational requirements. This approach achieves accurate sensitivity estimation while completing analyses in minutes rather than days for complex biological models [54].

Multivariate GSA Techniques: Complex models often produce multivariate outputs (e.g., temporal trajectories or multiple response variables). Novel sensitivity measures based on optimal transport theory enable comprehensive GSA for such systems by treating multivariate outputs as single entities, preserving their covariance structure during analysis [55].

Methods for Correlated Inputs: BLSS models often contain physiologically correlated parameters. Recent GSA methods based on copula theory and optimal transport can handle dependent inputs, providing meaningful sensitivity indices without requiring parameter independence [55].

Computational Methods and Protocols

SMoRe GloS Framework for Computationally Expensive Models

The SMoRe GloS methodology provides a structured, efficient approach to GSA for complex biological models through five systematic steps [54]:

Step 1: Generate ABM Output Sample parameter values across the defined parameter space Ω using structured sampling techniques. Latin Hypercube Sampling (LHS) or low-discrepancy sequences (e.g., Sobol sequences) provide better space-filling properties than random sampling. For each sampled parameter vector, run multiple BLSS model simulations to capture stochastic variability and compute averaged behavioral outputs.

Step 2: Formulate Candidate Surrogate Models Develop simplified mathematical representations (surrogate models) informed by the underlying biological mechanisms of the BLSS. For mass flow modeling, candidates might include simplified stoichiometric networks, response surface models, or reduced-order mechanistic models. Selection should be guided by the specific output metrics of interest.

Step 3: Select Optimal Surrogate Model Fit each candidate surrogate model to the BLSS output data using maximum likelihood estimation or weighted least squares optimization. Evaluate models using goodness-of-fit criteria (e.g., AIC, BIC, R²) and parameter identifiability metrics. Select the model that best balances accuracy and simplicity.

Step 4: Infer Relationship Between Surrogate and BLSS Parameters Establish mathematical relationships between BLSS parameters and surrogate model parameters using regression techniques, correlation analysis, or machine learning methods. Quantify uncertainty in these relationships through confidence interval estimation.

Step 5: Compute Global Sensitivity Indices Calculate sensitivity indices using the efficient surrogate model instead of the computationally expensive full BLSS model. Apply variance-based methods (e.g., eFAST, Sobol indices) or moment-independent approaches to the surrogate to obtain accurate sensitivity estimates for the original BLSS parameters.

Table 1: Comparison of Global Sensitivity Analysis Methods

Method	Key Features	Computational Cost	Best Use Cases	Limitations
Morris (MOAT)	One-at-a-time screening method	Low	Factor prioritization, preliminary screening	Limited interaction analysis, qualitative rankings only
Sobol Indices	Variance-based, full interaction quantification	High	Comprehensive analysis, factor fixing	Requires specialized sampling, computationally intensive
eFAST	Variance-based, efficient Fourier analysis	Medium-High	Factor prioritization, interaction assessment	Complex implementation, limited to variance-based measures
SMoRe GloS	Surrogate-based, compatible with various methods	Low (after surrogate built)	Computationally expensive models, complex systems	Requires surrogate formulation, additional validation needed
Optimal Transport-based	Handles multivariate outputs, correlated inputs	Medium	Complex outputs, dependent parameters	Emerging method, limited software implementation

Experimental Protocol: GSA for BLSS Mass Flow Parameters

Materials and Reagents

• BLSS stoichiometric model (computational)
• High-performance computing resources
• Parameter distributions (empirical or literature-based)
• Sensitivity analysis software (SALib, Python, R)

Procedure

• Parameter Selection and Range Definition

  • Identify all uncertain parameters in the BLSS mass flow model (e.g., reaction rates, uptake constants, stoichiometric coefficients)
  • Define plausible ranges for each parameter based on experimental data or literature values
  • Establish probability distributions for each parameter (uniform, normal, log-normal)

• Sampling Design Generation

  • Generate parameter samples using Latin Hypercube Sampling (LHS) with space-filling properties
  • For k parameters, generate N samples where N > 500 × k for reliable sensitivity estimates
  • Include edge samples and interior points for comprehensive space coverage

• Model Evaluation

  • Execute BLSS model for each parameter sample in the design matrix
  • Run multiple replicates for stochastic models to account for inherent variability
  • Record all relevant output metrics (mass flows, equilibrium states, stability indicators)

• Sensitivity Index Calculation

  • Compute first-order and total-order Sobol indices using the model output
  • Apply bootstrap resampling (≥1000 iterations) to estimate confidence intervals
  • Calculate interaction effects through higher-order indices or difference measures (STi - Si)

• Result Interpretation and Validation

  • Rank parameters by sensitivity indices to identify critical factors
  • Validate results through comparison with alternative GSA methods (e.g., Morris method)
  • Perform statistical tests to confirm significance of sensitivity rankings

Expected Outcomes

• Quantified sensitivity indices for all BLSS parameters
• Identification of 3-5 most critical parameters driving mass flow uncertainty
• Understanding of interaction effects between key parameters
• Guidance for targeted parameter estimation and experimental design

Implementation and Visualization

Research Reagent Solutions

Table 2: Essential Computational Tools for GSA Implementation

Tool/Category	Specific Examples	Function in GSA Workflow	Implementation Considerations
Sampling Algorithms	Latin Hypercube, Sobol sequences, Monte Carlo	Generate parameter combinations for model evaluation	Balance between space-filling properties and sample size
Variance-Based Methods	Sobol indices, eFAST, PAWN	Quantify main and interaction effects	Computational cost increases exponentially with parameters
Screening Methods	Morris method, Derivative-based	Preliminary factor prioritization	Efficient for models with many parameters (>50)
Software Packages	SALib (Python), sensitivity (R), UQLab	Implement various GSA methods	SALib provides open-source, well-documented implementation
Surrogate Models	Polynomial chaos, Gaussian processes, SMoRe GloS	Approximate complex model behavior	Balance between accuracy and computational efficiency
Visualization Tools	Tornado plots, Spider charts, Heat maps	Communicate sensitivity results	Prioritize clarity in presenting multidimensional relationships

Workflow Visualization

The following diagram illustrates the complete SMoRe GloS workflow for BLSS parameter sensitivity analysis:

GSA Workflow for BLSS Models

Sensitivity Results Communication

Effective visualization of GSA results enables clear interpretation of complex sensitivity patterns:

Sensitivity Results Interpretation

Application to BLSS Stoichiometric Modeling

Critical Parameter Identification in Mass Flow Networks

Implementing GSA for BLSS stoichiometric models enables researchers to identify which biological and physical parameters most significantly influence critical system outputs. For mass flow modeling, key parameters typically include plant growth rates, nutrient uptake efficiencies, microbial respiration rates, waste processing kinetics, and gas exchange coefficients. Through systematic GSA, researchers can determine which parameters require precise estimation and which have negligible impact on system performance.

Application of variance-based GSA methods to BLSS models has revealed that approximately 70-80% of output variance typically derives from 20-30% of input parameters, following a Pareto-like distribution [54] [55]. This pattern enables focused research efforts on the most influential parameters, optimizing resource allocation in experimental characterization studies.

Protocol for BLSS-Specific GSA

Specialized Considerations for BLSS Applications:

• Temporal Dynamics Analysis: BLSS models produce time-dependent outputs. Implement GSA at multiple time points or use multivariate methods that preserve temporal structure.

• Correlated Parameter Handling: Biological parameters in BLSS often exhibit physiological correlations. Employ GSA methods robust to parameter dependencies, such as those based on copula theory or optimal transport.

• Multi-Output Optimization: BLSS performance involves balancing multiple objectives (O₂ production, CO₂ consumption, food production, water purification). Use multivariate GSA techniques that accommodate multiple response variables simultaneously.

• Experimental Validation: Design targeted experiments to refine estimates of high-sensitivity parameters identified through GSA, creating an iterative model improvement cycle.

Expected Benefits for BLSS Research:

• Reduced uncertainty in system performance predictions
• Optimized experimental design for parameter estimation
• Improved system stability through better understanding of critical control points
• Enhanced fault detection and mitigation strategies

The integration of robust GSA methodologies into BLSS stoichiometric modeling represents a powerful approach for enhancing system reliability and guiding efficient research resource allocation. By identifying the parameters that matter most, researchers can focus their efforts where they will have the greatest impact on system performance and mission success.

Robust Model Validation, Selection, and Comparative Analysis

The χ2-Test of Goodness-of-Fit in Metabolic Flux Analysis (13C-MFA)

In the context of modeling mass flows in Bioregenerative Life Support Systems (BLSS), the accurate quantification of intracellular metabolic fluxes is paramount. 13C Metabolic Flux Analysis (13C-MFA) has emerged as the gold standard technique for this purpose, providing unparalleled insights into cellular physiology by quantifying in vivo reaction rates within metabolic networks [56] [57]. The reliability of flux estimates obtained through 13C-MFA hinges critically on proper statistical validation, with the χ2-test of goodness-of-fit serving as a cornerstone for evaluating model adequacy [58] [59]. This test determines whether the mathematical model of the metabolic network provides a statistically adequate fit to the experimental isotopic labeling data, thereby ensuring that subsequent flux interpretations are biologically meaningful [58].

The application of the χ2-test within 13C-MFA represents a critical gatekeeping function in the iterative process of model development and refinement. As researchers work to reconcile complex metabolic network structures with precise mass isotopomer distribution (MID) measurements, the χ2-test provides the quantitative rigor necessary to distinguish between physiologically plausible flux maps and those that fail to capture the underlying metabolic state [59]. This is particularly crucial in BLSS research, where understanding metabolic partitioning and carbon conversion efficiency directly impacts system design and organism selection.

Theoretical Foundation of the χ2-Test in 13C-MFA

Mathematical Formulation

In 13C-MFA, the χ2-test is implemented as a weighted least-squares optimization problem where the objective is to minimize the difference between experimentally observed and model-simulated mass isotopomer distributions [60] [58]. The fundamental mathematical formulation involves calculating the residual sum of squares (RSS) between measured and simulated data points:

The goodness-of-fit statistic is computed as: [ \chi^2 = \sum{i=1}^{n} \frac{(Oi - Ei)^2}{\sigmai^2} ] where (Oi) represents the observed measurement, (Ei) is the model-predicted value, (\sigma_i) is the standard deviation of the measurement, and (n) is the total number of measurements [58].

The resulting χ2 value follows a χ2-distribution with degrees of freedom (df) equal to: [ df = n - p ] where (p) represents the number of independently adjusted flux parameters in the model [58]. A model is considered statistically adequate if the computed χ2 value is less than the critical χ2 value at a chosen significance level (typically α = 0.05) [58].

Key Assumptions and Requirements

The validity of the χ2-test in 13C-MFA depends on several critical assumptions:

• Measurement errors are independent and normally distributed [58]
• Error magnitudes (σ) are accurately known [58]
• Metabolic steady-state is maintained throughout the labeling experiment [60]
• Isotopic mass effects are negligible [61]

In practice, the accurate determination of measurement errors (σ) presents a significant challenge, as underestimation can lead to model rejection even for correct models, while overestimation may result in the acceptance of incorrect models [58].

Workflow Integration and Protocol Implementation

Comprehensive 13C-MFA Workflow with χ2-Test Integration

The following diagram illustrates the complete 13C-MFA workflow, highlighting the central role of the χ2-test in model validation:

Figure 1: 13C-MFA workflow showing χ2-test integration for model validation.

Detailed Protocol for χ2-Test Implementation

Step 1: Experimental Design and Data Collection

• Tracer Selection: Choose appropriate 13C-labeled substrates (e.g., [1-13C]glucose, [U-13C]glucose) based on the metabolic pathways of interest [57] [61].
• Culturing Conditions: Maintain cells in metabolic steady-state using chemostat or steady-state batch cultures [60].
• Mass Isotopomer Measurement:
  • Harvest cells during isotopic steady-state
  • Extract intracellular metabolites
  • Analyze via GC-MS or LC-MS to obtain mass isotopomer distributions (MIDs)
  • Calculate measurement errors from biological replicates [56] [58]

Step 2: Metabolic Network Model Preparation

• Stoichiometric Matrix: Construct stoichiometric matrix (S) representing all metabolic reactions
• Atom Transitions: Define carbon atom transitions for each reaction
• Free Fluxes: Identify free flux parameters to be estimated
• Measurement Vector: Compile all measured MIDs and external fluxes [60] [61]

Step 3: Flux Estimation and χ2-Test Execution

• Parameter Estimation: Solve the nonlinear optimization problem to find flux parameters (v) that minimize the residuals between measured and simulated MIDs [58]
• Goodness-of-Fit Calculation:

• Statistical Evaluation:
  • Determine degrees of freedom: df = nmeasurements - nparameters
  • Obtain critical χ2 value from χ2-distribution table (typically α = 0.05)
  • Compare calculated χ2 to critical value [58]

Step 4: Interpretation and Decision

• Adequate Fit: If χ2calculated < χ2critical, the model is statistically adequate
• Poor Fit: If χ2calculated > χ2critical, investigate potential causes [58]

Troubleshooting and Advanced Applications

Common Issues and Solutions

Table 1: Troubleshooting χ2-test problems in 13C-MFA

Problem	Potential Causes	Diagnostic Approaches	Solutions
Persistent poor fit (high χ2 value)	Incorrect network topology [58]	Compare alternative models using validation data [58]	Add missing reactions or compartments
	Underestimated measurement errors [58]	Analyze error distributions from replicates	Adjust error estimates or use validation-based approach [58]
	Metabolic non-steady-state [60]	Check labeling time courses	Use INST-MFA instead of steady-state MFA [60]
Overfitting (unrealistically low χ2)	Overestimated measurement errors [58]	Compare χ2 values across models with different complexities	Use stricter error estimates or validation data [58]
	Excessive model complexity [59]	Perform model selection with independent data [58]	Apply parsimonious model selection
Unidentifiable fluxes	Insufficient labeling measurements [61]	Analyze flux confidence intervals	Implement parallel labeling experiments [61]

Advanced Implementation: Validation-Based Model Selection

Recent advances address limitations of the traditional χ2-test by incorporating validation-based model selection [58]. This approach uses independent validation data not used in model fitting, providing robustness against measurement error uncertainty:

Figure 2: Validation-based model selection workflow as an advanced alternative.

Research Reagent Solutions and Computational Tools

Table 2: Essential research reagents and computational tools for 13C-MFA

Category	Specific Examples	Function/Application	Implementation Considerations
Isotopic Tracers	[1-13C]Glucose, [U-13C]Glucose [57]	Reveals specific pathway activities	Selection depends on pathways of interest; purity >99% required
	13C-Acetate, 13C-Glutamine [60]	Studies TCA cycle and anaplerotic fluxes	Cost optimization through parallel labeling designs [61]
Analytical Instruments	GC-MS systems [56]	Measures mass isotopomer distributions	Requires proper calibration and natural abundance correction
	LC-MS/MS systems [60]	Enhanced measurement precision	Higher sensitivity for low-abundance metabolites
Software Platforms	OpenFLUX2 [61]	Open-source flux estimation	Supports parallel labeling experiments
	13CFLUX2 [61]	Comprehensive flux analysis	Implements EMU framework for computational efficiency
	Metran [57]	Isotopic steady-state MFA	Includes statistical evaluation tools
Statistical Packages	MATLAB-based tools [58]	Custom model development	Flexible but requires programming expertise
	R packages for χ2 analysis [59]	Goodness-of-fit testing	Open-source alternative to commercial software

The χ2-test of goodness-of-fit remains an indispensable component of rigorous 13C-MFA, providing the statistical foundation for validating metabolic network models and ensuring the reliability of estimated flux maps. When properly implemented within a comprehensive workflow that includes careful experimental design, appropriate model selection, and thorough statistical evaluation, the χ2-test serves as a critical checkpoint for flux analysis in BLSS research and related fields. The emergence of validation-based approaches complements traditional χ2-testing, offering enhanced robustness against measurement uncertainty and facilitating the development of more accurate metabolic models for understanding and engineering biological systems.

Validating FBA Predictions Against Experimental Flux Maps

Flux Balance Analysis (FBA) is a cornerstone computational method in systems biology that predicts steady-state metabolic fluxes in biochemical networks [34]. Unlike 13C-Metabolic Flux Analysis (13C-MFA), which estimates fluxes from experimental isotopic labeling data, FBA uses linear optimization to predict flux distributions based on stoichiometric constraints and assumed cellular objectives [62] [59]. Validating these predictions against experimental flux maps is crucial for ensuring biological relevance and enhancing predictive accuracy, particularly in specialized applications such as Bioregenerative Life Support Systems (BLSS) where reliable metabolic simulations are essential for long-duration space missions [1]. This protocol outlines comprehensive methodologies for rigorously validating FBA-derived flux predictions using experimental data, enabling researchers to quantify confidence in model outputs and refine model architectures for improved biological fidelity.

Theoretical Foundations of FBA Validation

Fundamental Principles of Flux Balance Analysis

FBA operates on the principle of mass balance constraint at metabolic steady state, where the stoichiometric matrix (S) defines the network structure and the relationship between metabolites and reactions [34] [63]. The fundamental equation, S·v = 0, represents the steady-state condition where metabolite concentrations remain constant over time. FBA extends this by optimizing an objective function (typically biomass maximization or product synthesis) subject to additional physico-chemical constraints [62] [63]. The linear programming formulation becomes:

Maximize: Z = cᵀv Subject to: S·v = 0 vmin ≤ v ≤ vmax

where Z represents the objective function, c is the vector of coefficients defining the objective, and vmin/vmax represent lower/upper bounds on reaction fluxes [63]. This framework allows prediction of flux distributions but requires validation against experimental data to confirm biological relevance.

The Critical Role of Validation in Metabolic Modeling

Validation transforms FBA from a purely theoretical exercise to a biologically relevant modeling approach. Several critical issues necessitate robust validation protocols: (1) uncertainty in network reconstruction and gap-filling, (2) potential mis-specification of cellular objective functions, (3) inadequate representation of regulatory constraints, and (4) mathematical degeneracy where multiple flux distributions yield identical objective values [62] [64]. Without proper validation, FBA predictions may represent mathematical optima with little connection to biological reality, potentially leading to erroneous conclusions in both basic research and applied biotechnology [59].

Comprehensive Validation Approaches and Methodologies

Classification of Validation Techniques

Multiple complementary approaches exist for validating FBA predictions, each with distinct strengths and applications. These methods can be systematically categorized based on their underlying principles and data requirements.

Table 1: Classification of FBA Validation Approaches

Validation Category	Underlying Principle	Data Requirements	Key Applications	Limitations
Direct Flux Comparison	Quantitative comparison of FBA-predicted versus experimentally measured internal fluxes	13C-MFA flux maps for central metabolism [62]	Gold standard validation for core metabolic pathways	Limited to central metabolism; technically challenging
Phenotypic Growth Validation	Comparison of predicted vs. observed growth capabilities and rates	Growth phenotypes across multiple substrates/environments [59]	High-throughput model validation; essential quality control	Does not validate internal flux distributions
Objective Function Validation	Statistical identification of biologically relevant objective functions	Experimental flux data for training and validation [65] [64]	Improving model accuracy; understanding cellular priorities	Method-dependent results; computational complexity
Multi-Model Statistical Assessment	Comparison of alternative model architectures using goodness-of-fit tests	Comprehensive flux and labeling data [62] [66]	Model selection and refinement	Requires substantial experimental data

Detailed Validation Protocols

Protocol 1: Direct Validation Against 13C-MFA Flux Maps

Principle: This approach provides the most rigorous validation by comparing FBA-predicted intracellular fluxes against those estimated from 13C-MFA, which serves as an experimental reference [62].

Experimental Requirements:

• 13C-labeling data from cells grown in defined conditions matching FBA simulation parameters
• Mass spectrometry or NMR measurements of isotopic labeling patterns
• Flux estimation using specialized software (e.g., INCA, OpenFLUX)
• Quantification of flux uncertainties through statistical analysis

Procedure:

• Experimental Data Acquisition: Grow cells under controlled conditions with 13C-labeled substrates (e.g., [1-13C]glucose). Extract metabolites and measure mass isotopomer distributions using LC-MS or GC-MS [62].
• Flux Estimation: Use 13C-MFA software to estimate intracellular fluxes that best fit the experimental labeling data. Calculate confidence intervals for all flux estimates [62].
• FBA Simulation: Perform FBA simulations using identical nutritional and environmental conditions as in the labeling experiment.
• Quantitative Comparison: Calculate similarity metrics between FBA-predicted and MFA-estimated fluxes:
  • Pearson correlation coefficient (R)
  • Mean absolute error (MAE)
  • Normalized root mean square error (NRMSE)
• Statistical Assessment: Evaluate whether FBA predictions fall within confidence intervals of MFA estimates for key central metabolic fluxes.

Interpretation: Strong correlation (R > 0.9) with minimal deviation (NRMSE < 0.2) indicates high model accuracy. Systematic discrepancies suggest model gaps or incorrect objective functions [62].

Protocol 2: Phenotypic Growth Validation

Principle: This method validates FBA models by comparing predicted growth capabilities (qualitative) and rates (quantitative) against experimental measurements across multiple conditions [59].

Procedure:

• Qualitative Growth/No-Growth Assessment:
  • Compile experimental data on growth capabilities across different carbon sources.
  • Perform FBA simulations for each condition, testing growth prediction accuracy.
  • Calculate Matthews Correlation Coefficient (MCC) to quantify predictive performance for binary growth outcomes.

• Quantitative Growth Rate Comparison:

  • Measure experimental growth rates in defined media conditions.
  • Compare against FBA-predicted growth rates.
  • Calculate R² values and absolute errors between predicted and observed rates.

• Condition Transfer Test:

  • Validate models by predicting growth in conditions not used during model construction.
  • Assess generalizability beyond training data.

Applications: Essential for initial model quality control and high-throughput validation [59].

Protocol 3: Advanced Objective Function Validation Using BOSS/TIObjFind

Principle: This approach identifies the most biologically relevant objective function by minimizing differences between predicted and experimental fluxes, addressing a key uncertainty in FBA [65] [64].

Procedure:

• Experimental Flux Data Collection: Obtain reliable flux measurements (e.g., from 13C-MFA) for training.
• BOSS Framework Implementation:
  • Formulate as optimization problem minimizing sum-squared error between experimental and predicted fluxes.
  • Identify putative objective reaction coefficients through dual optimization.
  • Add novel objective reactions to stoichiometric matrix when necessary [65].
• TIObjFind Framework Application:
  • Integrate Metabolic Pathway Analysis (MPA) with FBA.
  • Calculate Coefficients of Importance (CoIs) quantifying each reaction's contribution to objective function.
  • Construct flux-dependent weighted reaction graph [64].
• Cross-Validation: Validate identified objective functions with independent flux data not used during training.

Output: Statistically justified objective function with improved predictive accuracy for internal flux distributions [64].

Implementation Workflow and Visualization

The validation of FBA predictions follows a systematic workflow that integrates computational and experimental components. This multi-stage process ensures comprehensive assessment of model performance and biological relevance.

Figure 1: Comprehensive workflow for validating FBA predictions against experimental data, showing the sequential process from model preparation through validation to iterative refinement. The workflow integrates multiple validation methodologies that can be applied individually or in combination based on research objectives and data availability.

Essential Research Reagents and Computational Tools

Successful implementation of FBA validation requires both experimental reagents and computational resources. The following table details essential components for executing the validation protocols described in this document.

Table 2: Essential Research Reagent Solutions and Computational Tools

Category	Item/Resource	Specification/Purpose	Application Notes
Isotopic Tracers	[1-13C]Glucose	Carbon labeling for 13C-MFA; >99% isotopic purity	Enables flux estimation in central carbon metabolism [62]
Analytical Instruments	LC-MS/MS or GC-MS Systems	Mass isotopomer distribution measurement	Required for experimental flux determination [62]
Cell Culture Components	Defined Growth Media	Chemically defined formulation without uncharacterized components	Essential for controlled FBA validation studies [59]
Computational Tools	COBRA Toolbox	MATLAB-based FBA simulation environment	Standard platform for constraint-based modeling [59]
Computational Tools	BOSS/TIObjFind	Objective function identification algorithms	Advanced objective function validation [65] [64]
Computational Tools	MEMOTE Suite	Metabolic model testing and quality control	Automated model validation and consistency checking [59]
Data Resources	BiGG Models Database	Curated genome-scale metabolic models	Reference models for validation studies [59]

Application to BLSS Stoichiometric Modeling

In Bioregenerative Life Support Systems (BLSS), validating metabolic models takes on additional significance due to the critical nature of these systems for long-duration space missions [1]. BLSS implementations, such as the MELiSSA system, require precise stoichiometric modeling of mass flows through interconnected compartments containing microorganisms, plants, and humans [1]. FBA validation in this context presents unique challenges and considerations:

System-Level Validation Requirements: BLSS models must maintain element cycling (C, H, O, N) across multiple species while ensuring complete closure of mass flows [1] [67]. Validation must therefore extend beyond single-organism metabolism to encompass cross-compartment flux consistency.

Multi-Scale Validation Approach: Effective BLSS model validation requires:

• Component-Level Validation: Individual organism FBA models validated against experimental data
• Subsystem Validation: Key processes (e.g., waste degradation, oxygen production) validated independently
• Integrated System Validation: Overall system closure and stability assessment [1]

Closure Metrics: For BLSS applications, successful validation should demonstrate minimal loss of critical elements between system iterations, with ideal performance showing zero loss for most compounds and only minor losses for gases like O₂ and CO₂ [1].

Robust validation of FBA predictions against experimental flux maps is essential for enhancing confidence in constraint-based modeling and expanding its applications in biotechnology and systems biology [62] [59]. The protocols outlined here provide a comprehensive framework for assessing model accuracy, from basic phenotypic validation to advanced objective function identification. As the field progresses, several emerging trends promise to further strengthen validation practices: (1) integration of multi-omic data for comprehensive model constraints, (2) development of automated validation pipelines, and (3) implementation of machine learning approaches to identify patterns in validation discrepancies. For BLSS applications and other critical biotechnologies, rigorous validation is not merely an academic exercise but a necessary step toward reliable implementation of metabolic models in engineering biological systems.

Bayesian Techniques for Characterizing Flux Uncertainty

Quantifying uncertainty is paramount in stoichiometric modeling of Biological Life Support System (BLSS) mass flows, where accurate predictions of element fluxes are essential for system stability and reliability. Bayesian methods provide a powerful probabilistic framework for characterizing this uncertainty, moving beyond single-point estimates to deliver full probability distributions for model parameters and predictions [68]. This approach formally integrates prior knowledge with experimental data, offering a robust mechanism for updating beliefs about mass fluxes as new information becomes available from BLSS operations or related experiments [69]. The adoption of Bayesian techniques aligns with the iterative nature of BLSS development, allowing models to learn sequentially from successive experimental campaigns and operational data, thereby refining uncertainty estimates over time [70].

Theoretical Foundations of Bayesian Flux Analysis

Core Bayesian Principles for Flux Estimation

Bayesian methods treat unknown parameters as random variables characterized by probability distributions. This paradigm is built upon Bayes' Theorem, which updates prior beliefs about parameters with observed data. For flux analysis, the theorem is expressed as:

P(Fluxes | Data) = [ P(Data | Fluxes) × P(Fluxes) ] / P(Data)

where:

• P(Fluxes | Data) is the posterior distribution, representing updated belief about fluxes after observing data
• P(Data | Fluxes) is the likelihood function, indicating how probable the observed data are for different flux values
• P(Fluxes) is the prior distribution, encapsulating pre-existing knowledge about possible flux values
• P(Data) is the marginal likelihood, serving as a normalizing constant [70] [71]

This framework is particularly valuable for BLSS applications where prior information may exist from previous experimental campaigns, terrestrial analogs, or theoretical considerations. The Bayesian approach naturally accommodates the complex, integrated nature of BLSS mass flows, where multiple biological and physicochemical processes interact within materially closed systems [16].

Advantages Over Traditional Methods for BLSS Applications

Traditional frequentist approaches to flux analysis often rely on point estimates and confidence intervals based on hypothetical repeated experiments. In contrast, Bayesian methods provide direct probability statements about fluxes, which is more intuitive for decision-making in BLSS design and operation [72]. The Bayesian framework readily handles complex models with multiple parameters, properly propagating uncertainty from all sources to provide comprehensive uncertainty quantification for predictions [71]. This is particularly important for BLSS, where predictions of future system states based on current measurements must include realistic uncertainty bounds to inform resource management and contingency planning.

Table 1: Comparison of Flux Analysis Approaches

Feature	Traditional Optimization	Bayesian Approach
Uncertainty Output	Confidence intervals	Full probability distributions
Prior Information	Difficult to incorporate	Naturally incorporated
Complex Models	Often computationally challenging	Handled with MCMC sampling
Result Interpretation	Based on hypothetical repeats	Direct probability statements
Sequential Learning	Requires specialized methods	Built into the framework

Computational Implementation

Markov Chain Monte Carlo Methods

Practical implementation of Bayesian flux analysis typically relies on Markov Chain Monte Carlo sampling methods, which enable numerical approximation of posterior distributions for complex models where analytical solutions are intractable. MCMC algorithms generate samples from the posterior distribution, creating an empirical approximation that can be used for inference and prediction [68]. The development of efficient MCMC algorithms, particularly Hamiltonian Monte Carlo, has dramatically improved the feasibility of Bayesian analysis for complex flux models [68]. Open-source probabilistic programming tools such as Stan provide accessible platforms for implementing these algorithms without requiring deep expertise in computational statistics [68].

Workflow for Bayesian Flux Analysis

The following diagram illustrates the iterative process of Bayesian flux analysis for BLSS applications:

Application to BLSS Stoichiometric Modeling

Mass Balance Formulation with Uncertainty

In BLSS stoichiometric modeling, mass balances form the foundation for predicting element fluxes through system components. A generalized mass balance for a BLSS can be represented as:

Accumulation = Input - Output + Generation - Consumption

Bayesian methods enhance this deterministic framework by treating each term as a probability distribution rather than a fixed value [16]. For example, in modeling carbon fluxes through a BLSS, the assimilation rates by plants, respiration rates by various organisms, and mass transfer between subsystems can all be represented as probability distributions that reflect their inherent variability and measurement uncertainty [73]. This approach naturally accommodates the biological variability inherent in living systems, providing more realistic uncertainty bounds on predictions of system behavior.

Case Study: Carbon Flux Uncertainty in Plant Growth Chambers

A representative application of Bayesian methods involves quantifying carbon dioxide assimilation uncertainty in BLSS plant growth modules. The following protocol outlines the experimental and computational approach:

Protocol 1: Carbon Flux Uncertainty Quantification

• Experimental Data Collection

  • Measure CO₂ concentrations at chamber inlet and outlet at 5-minute intervals over multiple light-dark cycles
  • Quantify air flow rates through the chamber with calibrated mass flow meters
  • Record photosynthetic photon flux density, temperature, and relative humidity concurrently
  • Destructively harvest plant biomass at beginning and end of experimental period to determine growth rates

• Flux Calculation

  • Compute instantaneous net CO₂ exchange rates using: Flux = FlowRate × (CO₂in - CO₂out)
  • Separate net photosynthesis and dark respiration rates based on light period measurements

• Model Specification

  • Define likelihood function: Data ~ Normal(PredictedFlux, σ)
  • Specify prior distributions for key parameters:
    • Maximum photosynthetic rate: Vmax ~ Normal(35, 10) μmol/m²/s
    • Respiration rate: Rd ~ Normal(2, 1) μmol/m²/s
    • Measurement error: σ ~ HalfNormal(5)
  • Implement Michaelis-Menten light response function for photosynthesis

• Bayesian Computation

  • Implement model in probabilistic programming language (e.g., Stan, PyMC)
  • Run 4 parallel MCMC chains with 2000 iterations each
  • Verify chain convergence (R̂ < 1.05)
  • Check posterior predictive distributions against experimental data

• Uncertainty Analysis

  • Extract posterior distributions for all parameters
  • Compute credible intervals for CO₂ assimilation rates under different light conditions
  • Propagate uncertainty to predict oxygen production rates
  • Perform sensitivity analysis on prior distribution choices

Table 2: Key Research Reagents and Computational Tools for Bayesian Flux Analysis

Item	Function	Example Sources/Platforms
Probabilistic Programming Languages	Model specification and inference	Stan, PyMC, Turing.jl
MCMC Samplers	Posterior distribution computation	NUTS, HMC, Metropolis
Convergence Diagnostics	Verifying sampling quality	R̂, effective sample size, trace plots
Gas Analyzers	Concentration measurements	IRGA, mass spectrometry
Flow Measurement	Volumetric or mass flow rates	Coriolis flow meters, mass flow controllers
Data Logging Systems	Experimental data acquisition	Custom LABVIEW, Python, or R scripts

Advanced Protocol: Bayesian Calibration of Stoichiometric Models

For comprehensive BLSS modeling, Bayesian methods can calibrate entire stoichiometric networks using multiple data types. The following protocol extends the approach to multi-element mass balances:

Protocol 2: Multi-Element Stoichiometric Model Calibration

• Stoichiometric Matrix Formulation

  • Define system components and processes
  • Create stoichiometric matrix encoding mass balance relationships
  • Identify measurable and unmeasurable fluxes

• Prior Information Elicitation

  • Conduct literature review for prior distributions on biological conversion rates
  • Incorporate thermodynamic constraints as informative priors
  • Use BLSS closure requirements to inform flux boundaries

• Hierarchical Model Specification

  • Implement hierarchical structure to share information across similar processes
  • Specify appropriate likelihood functions for different measurement types
  • Include covariance structure for correlated fluxes

• Model Checking and Validation

  • Perform posterior predictive checks
  • Compare with hold-out validation data
  • Conduct cross-validation across multiple experimental runs

The following diagram illustrates the relationship between model components, data sources, and uncertainty in a Bayesian stoichiometric network analysis:

Reporting Guidelines for Bayesian Flux Analysis

Transparent reporting of Bayesian analyses is essential for reproducibility and scientific credibility. The following key elements should be documented in BLSS flux studies:

• Prior Distributions: Clearly specify all prior distributions used, including their justification based on BLSS literature, theoretical constraints, or previous experiments [72].

• Computational Details: Report the MCMC algorithm used, number of chains, iterations, burn-in period, and convergence diagnostics [72].

• Sensitivity Analysis: Demonstrate how results change with different prior specifications, particularly for potentially contentious prior choices [72].

• Posterior Summaries: Present appropriate summaries of posterior distributions, including measures of central tendency and credible intervals [72].

• Model Code and Data: Make analysis code and processed data available to enable verification and extension of the work [72].

Adhering to these guidelines ensures that Bayesian analyses of BLSS mass flows can be properly evaluated, compared, and built upon by the research community, accelerating progress toward reliable closed-loop life support systems for long-duration space missions.

Parallel Labeling Experiments for Improved Flux Resolution

Parallel Labeling Experiments (PLEs) represent a robust methodology in metabolic flux analysis (MFA) where multiple isotopic tracer experiments are conducted simultaneously under identical biological conditions, varying only the labeling pattern of the administered substrates [74] [75]. This approach stands in contrast to single labeling experiments, providing a synergistic effect that significantly enhances the resolution and precision of calculated metabolic fluxes. The fundamental principle underpinning PLEs is that different isotopic tracers illuminate distinct segments of the metabolic network. By integrating data from these complementary experiments, researchers can achieve a more comprehensive and accurate quantification of intracellular reaction rates, a critical requirement for advanced research in Bioregenerative Life Support Systems (BLSS) where understanding mass flows is essential for system stability and efficiency [76] [77].

The transition from single to parallel labeling strategies marks a significant evolution in fluxomics. Historically, metabolic pathway elucidation relied heavily on radioisotopes like 14C and 3H, where parallel experiments with different radioactive tracers were used to target specific pathways [74]. The advent of stable isotopes (e.g., 13C, 15N), coupled with advances in analytical technologies such as Mass Spectrometry (MS) and Nuclear Magnetic Resonance (NMR) spectroscopy, has facilitated more complex and informative labeling strategies [74] [78]. The COMPLETE-MFA methodology, which is founded on the integrated analysis of multiple parallel labeling experiments, has emerged as a gold standard in the field, demonstrating that PLEs can drastically improve flux observability and reduce confidence intervals, even for challenging exchange fluxes [76] [77].

Theoretical Foundations and Advantages

Core Principle: Information Synergy

The enhanced flux resolution achieved through PLEs stems from the synergistic information gained from complementary tracers. Each unique tracer provides specific information about the activity of different metabolic routes. For instance, in a study on E. coli, tracers such as 80% [1-13C]glucose + 20% [U-13C]glucose were optimal for resolving fluxes in the upper part of metabolism (glycolysis and pentose phosphate pathways), whereas [4,5,6-13C]glucose and [5-13C]glucose provided superior resolution for the lower part of metabolism (TCA cycle and anaplerotic reactions) [77]. No single tracer can optimally resolve all fluxes in a network; PLEs overcome this limitation by combining the strategic strengths of multiple tracers [77].

Key Advantages of the Parallel Approach

• Tailored Flux Resolution: PLEs can be designed to target and resolve specific, hard-to-measure fluxes with high precision, such as fluxes through parallel, cyclic, or bidirectional reversible reactions [74] [78].
• Validation of Network Models: The use of multiple tracers provides a built-in mechanism for validating the biochemical network model itself. Inconsistent model fits across different tracer datasets can indicate gaps or errors in the model structure [74] [75].
• Reduced Experimental Time: Introducing isotopes from multiple entry points can accelerate the propagation of labeling throughout the metabolic network, potentially shortening the time required for isotopes to reach isotopic steady state [74].
• Enhanced Performance in Data-Limited Systems: In systems where the number of measurable metabolites is constrained, PLEs compensate for the lack of extensive measurement data by providing a richer, more diverse labeling input dataset [75].

Experimental Protocol for COMPLETE-MFA

The following protocol outlines the core steps for implementing a parallel labeling experiment, based on the COMPLETE-MFA framework [76] [77].

Experimental Design and Tracer Selection

Objective: Identify a set of isotopic tracers that, when used in parallel, provide complementary information on the metabolic network.

• Define Metabolic Network: Construct a stoichiometric model of the central carbon metabolism, including atom transitions for each reaction.
• Identify Free Fluxes: Determine the number and location of independent fluxes (free fluxes) that define the flux state of the network.
• Select Tracer Combinations: Use rational design tools (e.g., EMU-based sensitivity analysis, robustified experimental design) to select optimal tracer mixtures [79] [80]. The goal is to find tracers that maximize the sensitivity of measured labeling patterns to changes in the free fluxes. For a foundational study, consider using all six singly labeled [1-13C] to [6-13C] glucose tracers [76].
• Plan Biological Replicates: A minimum of n=3 biological replicates per tracer condition is recommended to account for biological variability.

Cell Cultivation and Sampling

Objective: Grow cells in parallel cultures under metabolic steady-state conditions using the selected tracers.

• Prepare Labeled Media: Prepare identical culture media for each parallel experiment, where the sole carbon source is replaced with the respective 13C-labeled tracer. Ensure the same physiological state by starting all cultures from the same seed culture [74].
• Maintain Steady-State: For microbial cultures, use a chemostat to maintain a constant growth rate and metabolite concentrations. For mammalian cells, use controlled bioreactors. Verify steady-state by monitoring culture density, pH, and substrate levels over time.
• Harvest Samples: Once isotopic steady state is reached (typically after 3-5 residence times in a chemostat), harvest cells rapidly (e.g., via fast filtration) and quench metabolism immediately using liquid nitrogen or cold methanol-based solutions.

Metabolite Extraction and Analysis

Objective: Measure the mass isotopomer distributions (MIDs) of intracellular metabolites or proteinogenic amino acids.

• Metabolite Extraction: Use a validated extraction protocol (e.g., cold methanol/water followed by chloroform) to quench metabolism and extract polar intracellular metabolites.
• Derivatization: For Gas Chromatography-Mass Spectrometry (GC-MS) analysis, derive metabolites. A common target is proteinogenic amino acids from hydrolyzed biomass, as they provide stable, abundant labeling information that reflects the labeling of their central metabolic precursors.
• Mass Spectrometry Measurement: Analyze the derivatives using GC-MS. Measure the mass isotopomer distributions (MIDs), which represent the fractional abundances of molecules with different numbers of heavy isotopes.

Data Integration and Flux Calculation

Objective: Integrate labeling and extracellular flux data to compute the most probable flux map.

• Data Compilation: Compile the MIDs from all parallel experiments and extracellular uptake/secretion rates into a single dataset.
• Non-Linear Regression: Use a computational 13C-MFA software suite (e.g., 13CFLUX2) to fit the combined dataset to the metabolic model. The software performs an iterative least-squares regression to find the flux distribution that minimizes the difference between the simulated and measured MIDs.
• Statistical Analysis: Evaluate the goodness-of-fit (e.g., using chi-square tests) and calculate confidence intervals for the estimated fluxes (e.g., via Monte Carlo sampling or linear approximation).

The following workflow diagram illustrates the complete COMPLETE-MFA process, from experimental design to flux calculation.

Quantitative Tracer Performance Data

Selecting the right combination of tracers is critical to the success of a PLE campaign. The table below summarizes the performance of various glucose tracers in resolving fluxes in different parts of the E. coli metabolic network, as demonstrated in a large-scale study integrating 14 parallel labeling experiments [77].

Table 1: Performance of Glucose Tracers in Resolving Metabolic Fluxes in E. coli

Tracer Substrate	Optimal Flux Region	Key Strengths and Rationale
80% [1-13C]Glucose +20% [U-13C]Glucose	Upper Metabolism(Glycolysis, PPP)	Highly sensitive to pentose phosphate pathway split and glycolytic flux.
[4,5,6-13C]Glucose	Lower Metabolism(TCA Cycle, Anaplerosis)	Effectively traces carbon fate in pyruvate dehydrogenase, pyruvate carboxylase, and TCA cycle reactions.
[5-13C]Glucose	Lower Metabolism(TCA Cycle, Anaplerosis)	Provides distinct labeling patterns for TCA cycle intermediates, ideal for resolving malic enzyme and PEP carboxykinase fluxes.
[1,2-13C]Glucose	Glycolysis & PPP	Useful for quantifying reversible reactions in upper glycolysis and transaldolase/transketolase activities in the PPP.
[U-13C]Glucose	Global Network	Provides a global labeling baseline but may lack precision for specific, divergent pathways compared to optimized mixtures.

For mammalian cells, rational design approaches have identified novel optimal tracers. For instance, [2,3,4,5,6-13C]glucose is superior for elucidating the oxidative pentose phosphate (oxPPP) flux, while [3,4-13C]glucose is optimal for quantifying pyruvate carboxylase (PC) flux [79]. It is also demonstrated that 13C-glutamine tracers can perform poorly for these specific fluxes compared to the optimal glucose tracers [79].

Reagent and Tool Kit for Parallel Labeling Studies

A successful PLE study relies on a suite of specialized reagents and computational tools. The following table details the essential components of the scientist's toolkit for this methodology.

Table 2: Research Reagent and Tool Solutions for Parallel Labeling Experiments

Category / Item	Specification / Example	Function in Protocol
Stable Isotope Tracers	13C-labeled Glucose (e.g., [1-13C], [U-13C], [4,5,6-13C]); 13C-Glutamine	Serve as the isotopic source for tracing carbon fate through metabolic pathways. The core component of the experimental design.
Analytical Instrumentation	Gas Chromatography-Mass Spectrometry (GC-MS)	Workhorse for measuring mass isotopomer distributions (MIDs) of derivatized metabolites or proteinogenic amino acids.
Metabolic Modeling Software	13CFLUX2, INCA, OpenFLUX	High-performance software suites for simulating isotopic labeling and performing non-linear regression to calculate fluxes from labeling data.
Flux Modeling Language	FluxML	A universal model description language used to define the stoichiometric model, atom mappings, and experimental data [80].
Network Visualization & Editing	Omix Visualization Software	Facilitates the visual construction, editing, and validation of the metabolic network model before encoding in FluxML [80].

Pathway Diagram: Tracer Entry and Information Flow

The diagram below illustrates how different carbon atoms from variously labeled glucose tracers enter and propagate through the central metabolic network, highlighting the pathways illuminated by each tracer type. This visualizes the core concept of complementary information in PLEs.

Application in BLSS and Concluding Remarks

Within the context of Bioregenerative Life Support Systems (BLSS), the precise quantification of mass flows is not merely an academic exercise but a fundamental requirement for system modeling, optimization, and control. PLEs and the COMPLETE-MFA methodology provide an unparalleled toolset for achieving the high-resolution flux maps needed to understand and engineer the complex metabolic interactions within BLSS modules, whether they involve plant, microbial, or algal components. By accurately quantifying the carbon conversion efficiencies, nutrient recycling rates, and metabolic trade-offs between growth and maintenance, this approach directly informs the stoichiometric models of BLSS mass flows.

Future methodological developments will likely focus on addressing challenges such as managing biological variability across parallel cultures, standardizing data integration protocols, and further automating the rational design of tracer experiments to be more accessible for non-model organisms [74] [80]. The application of PLEs in BLSS research will be instrumental in transitioning from a qualitative understanding of metabolic capabilities to a quantitative, predictive science of mass and energy flows, thereby enhancing the reliability and sustainability of life support in long-duration space missions.

Comparative Analysis of Alternative Model Architectures and Objectives

Bioregenerative Life Support Systems (BLSS) are critical for enabling long-duration human space exploration by creating materially closed loops that regenerate air, water, and food from metabolic waste. These systems break down human waste materials into nutrients and CO₂ for plants and other edible organisms, which in turn provide food, fresh water, and oxygen for astronauts [1]. The central challenge lies in designing system architectures that achieve high degrees of material closure while maintaining operational stability and reliability.

Stoichiometric modeling provides the mathematical foundation for quantifying mass flows of carbon, hydrogen, oxygen, and nitrogen through these complex biological systems. This analysis examines alternative model architectures and their underlying objectives, with particular focus on the MELiSSA (Micro-Ecological Life Support System Alternative) framework developed by the European Space Agency and international partners [1]. The comparative assessment presented herein aims to guide researchers in selecting appropriate modeling approaches for specific BLSS development phases and mission requirements.

BLSS Model Architectures: Comparative Framework

Compartmentalized vs. Integrated Architectures

BLSS models employ distinct architectural approaches, primarily categorized as compartmentalized or integrated. Compartmentalized architectures, exemplified by the MELiSSA loop, separate biological processes into distinct interconnected modules, each inhabited by different types of organisms with specialized metabolic functions [1]. This modular approach facilitates system control, troubleshooting, and optimization of individual processes. In contrast, integrated architectures combine multiple biological processes within fewer compartments, potentially increasing stability through biological diversity but presenting challenges in process control and modeling.

Table 1: Comparative Analysis of BLSS Model Architectures

Architecture Type	Key Characteristics	Advantages	Limitations	Representative Systems
Compartmentalized	Discrete interconnected bioreactors; specialized organism functions	Enhanced controllability; simplified troubleshooting; predictable stoichiometry	Higher system complexity; inter-compartmental balancing challenges	MELiSSA [1]
Integrated	Combined biological processes; diverse microbial communities	Potential biological stability; reduced hardware requirements	Difficult process control; complex stoichiometric modeling	Early BLSS concepts [1]
Hybrid	Combines compartmentalization with redundant integrated elements	Balance of control and resilience; fault tolerance	Increased design complexity; larger mass/volume	Advanced MELiSSA variants [1]

MELiSSA Reference Architecture

The MELiSSA concept represents the most extensively developed compartmentalized architecture, consisting of five interconnected compartments inhabited by different types of organisms [1]. The system is designed to progressively break down organic waste through a sequence of biological processes:

• C1: Thermophilic anaerobic compartment performing liquefaction and initial decomposition
• C2: Photoheterotrophic compartment further breaking down organic materials
• C3: Nitrifying compartment converting ammonia to nitrates
• C4a & C4b: Photoautotrophic compartments (microalgae and higher plants) producing food and oxygen
• C5: Crew compartment (human habitation) generating waste inputs

This architecture creates a continuous metabolic loop where waste products from one compartment serve as resources for subsequent compartments, ultimately regenerating air, water, and food for the crew [1].

Stoichiometric Modeling Approaches

Modeling Objectives and Methodologies

Stoichiometric modeling of BLSS mass flows serves multiple objectives, from fundamental system design to operational control. These models describe the cycling of elements C, H, O, and N through all system compartments using balanced chemical equations with fixed or dynamically calculated coefficients [1].

Table 2: Stoichiometric Modeling Objectives and Methodologies

Modeling Objective	Primary Methodology	Element Coverage	Closure Target	Application Context
System Sizing	Fixed stoichiometric coefficients; steady-state assumption	C, H, O, N (essential); other minerals optional	<100% (with external inputs)	Preliminary mission design [1]
Dynamic Control	Dynamic coefficients; real-time parameter adjustment	C, H, O, N (comprehensive)	Variable (operational range)	Operational BLSS management [1]
Closure Analysis	Mass-balanced equations; element tracking	C, H, O, N (mandatory)	100% (theoretical target)	System validation [1] [81]
Reliability Assessment	Stochastic modeling; failure mode analysis	C, H, O, N (primary focus)	Degradation scenarios	Risk analysis [81]

Fully Closed vs. Partially Closed Models

A critical distinction in BLSS modeling approaches concerns the degree of material closure targeted. Traditional BLSS studies typically model systems where only a fraction of resources (such as food) are provided by the system itself, with the remainder supplied at mission initiation or through resupply [1]. In contrast, fully closed models aim for complete material recycling with minimal losses, which is essential for autonomous long-duration space missions without resupply possibilities [1] [81].

Recent advances in stoichiometric modeling have demonstrated the feasibility of achieving near-complete closure. The model developed by Vermeulen et al. achieved high closure at steady state, with 12 out of 14 compounds exhibiting zero loss, and only oxygen and CO₂ displaying minor losses between iterations [1]. This represents the first stoichiometric model of a MELiSSA-inspired BLSS that describes continuous provision of 100% of the food and oxygen needs of the crew [1].

Experimental Protocols for BLSS Model Validation

Stoichiometric Coefficient Determination Protocol

Objective: Quantify stoichiometric coefficients for mass flow equations through experimental measurement of biological processes in controlled bioreactors.

Materials:

• Bioreactors (anaerobic, photoheterotrophic, nitrifying, photoautotrophic configurations)
• Analytical instruments (HPLC, GC-MS, elemental analyzer, spectrophotometer)
• Sterile sampling equipment
• Data acquisition system
• Standard chemical solutions for calibration

Procedure:

• Inoculate each compartment with appropriate microbial consortia or plants
• Establish steady-state operation under controlled environmental conditions
• Implement continuous monitoring of input and output streams:
  • Gas composition (O₂, CO₂, CH₄, N₂)
  • Liquid phase nutrients (NO₃⁻, NH₄⁺, PO₄³⁻)
  • Biomass composition (elemental analysis)
  • Metabolic products (volatile fatty acids, dissolved organics)
• Collect triplicate samples at 24-hour intervals over 5 operational cycles
• Analyze samples using standardized analytical methods
• Calculate element mass balances for each compartment
• Derive stoichiometric coefficients through regression analysis of mass flow data
• Validate coefficients through independent batch experiments

Data Analysis:

• Compute elemental conversion efficiencies for each compartment
• Determine confidence intervals for stoichiometric coefficients
• Perform statistical analysis of mass closure (target >95% for all major elements)

System Closure Verification Protocol

Objective: Experimentally validate the degree of mass closure achieved in an integrated BLSS test facility.

Materials:

• Integrated BLSS pilot plant (e.g., MELiSSA Pilot Plant at Universitat Autònoma de Barcelona)
• Precision gas analysis system
• Liquid flow meters and sampling ports
• Mass spectrometers for isotope tracing
• Data logging infrastructure

Procedure:

• Establish baseline operation with crew analogue (physical/biological O₂ consumption, CO₂ production, waste generation)
• Implement continuous monitoring of all mass flows for 30-day operational period
• Introduce stable isotope tracers (¹³C, ¹⁵N) to track element pathways
• Measure accumulation of non-recyclable materials in waste streams
• Quantify process losses (degassing, precipitate formation, non-biodegradable fractions)
• Calculate mass closure metrics for each element:
  • Closure Efficiency = (1 - ΣLosses/ΣInputs) × 100%
  • Cycling Index = (Mass recycled/Total mass processed) × 100%
• Identify critical loss pathways and material bottlenecks
• Iterate system parameters to optimize closure metrics

Validation Criteria:

• Carbon closure >98% (excluding respiratory CO₂)
• Oxygen closure >99% (accounting for water electrolysis/formation)
• Nitrogen closure >99% (critical due to limited buffers)
• Water closure >99.9%

Visualization of BLSS Architectures and Modeling Approaches

MELiSSA Compartmentalized Architecture Workflow

Stoichiometric Model Development Process

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Research Reagents and Materials for BLSS Stoichiometric Modeling

Category	Specific Items	Function/Application	Critical Specifications
Analytical Standards	Certified gas mixtures (O₂, CO₂, CH₄, N₂); Ion chromatography standards (NO₃⁻, NH₄⁺, PO₄³⁻)	Instrument calibration; quantitative analysis	Certified reference materials; NIST-traceable
Microbial Cultures	Limnospira indica; Nitrifying bacteria consortia; Thermophilic anaerobes	Compartment inoculation; process validation	Axenic cultures; documented metabolic characteristics
Chemical Reagents	Stable isotopes (¹³C-glucose, ¹⁵N-urea); Elemental analysis standards; Digestion reagents	Tracer studies; biomass composition; sample preparation	Isotopic purity >99%; ACS grade reagents
Bioreactor Systems	Anaerobic chambers; Photobioreactors; Nitrification columns; Plant growth chambers	Process simulation; parameter optimization	Environmental control (T, pH, light); sampling ports
Analytical Instruments	Elemental analyzer; GC-MS; HPLC; Spectrophotometer; pH/conductivity meters	Composition analysis; concentration measurement	Appropriate detection limits; validated methods
Data Management	Laboratory Information Management System (LIMS); Process control software	Data integrity; experimental control	Audit trail capability; real-time monitoring

The comparative analysis of alternative model architectures and objectives for BLSS stoichiometric modeling reveals distinct trade-offs between complexity, controllability, and closure efficiency. Compartmentalized architectures like MELiSSA provide a structured framework for achieving high material closure through specialized biological processes, with demonstrated capability to approach 100% provision of food and oxygen needs for crewed missions [1]. The experimental protocols and visualization tools presented herein provide researchers with standardized methodologies for model development and validation. As humanity ventures toward long-duration space missions without resupply possibilities, these stoichiometric modeling approaches will become increasingly critical for mission success [81]. Future work should focus on dynamic modeling approaches that can accommodate system perturbations and long-term operational stability while maintaining high closure efficiencies.

Conclusion

Stoichiometric modeling provides a powerful, unifying framework for understanding and engineering complex biological systems, from closed-loop life support to human metabolism. The methodologies developed for BLSS, particularly the MELiSSA project, demonstrate that achieving near-complete mass closure is feasible through meticulous stoichiometric balancing and compartmental design. Concurrent advances in constraint-based modeling, such as FBA and ll-FBA, offer robust tools for predicting cellular behavior, while emerging techniques in machine learning and global sensitivity analysis are overcoming longstanding optimization hurdles. The rigorous validation practices established in 13C-MFA are crucial for building confidence in all flux predictions. The future of this field lies in further integrating these approaches, creating multi-scale models that can inform not only the design of sustainable ecosystems in space but also novel therapeutic strategies and a deeper understanding of metabolic diseases in clinical settings.

References

• 1. Stoichiometric model of a fully closed bioregenerative life ... [https://www.frontiersin.org/journals/astronomy-and-space-sciences/articles/10.3389/fspas.2023.1198689/full]
• 2. Critical investments in bioregenerative life support systems for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12357894/]
• 3. Plant and microbial science and technology as ... [https://www.nature.com/articles/s41526-023-00317-9]
• 4. A case for supporting human long-term survival on the moon [https://www.sciencedirect.com/science/article/abs/pii/S0094576524007343]
• 5. Synthesis and Evaluation of Bio-Lubricants from ... [https://www.mdpi.com/2075-4442/12/12/446]
• 6. Medical Operations Team Activities [https://humans-in-space.jaxa.jp/en/biz-lab/med-in-space/healthcare/medops/system/]
• 7. MELISSA: global control strategy of the artificial ecosystem ... [https://pubmed.ncbi.nlm.nih.gov/11542550/]
• 8. ESA - Closed Loop Compartments [https://www.esa.int/Enabling_Support/Space_Engineering_Technology/Melissa/Closed_Loop_Compartments]
• 9. Enhancing control systems of higher plant culture chambers ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9632494/]
• 10. 1.4.3 Empirical and Molecular Formulas [https://www.tutorchase.com/notes/ib/chemistry-2025-hl/1-4-3-empirical-and-molecular-formulas]
• 11. 1.1.5: Empirical Formulas from Analysis [https://chem.libretexts.org/Courses/City_College_of_San_Francisco/Chemistry_101A/01%3A_Topic_A-_Equations_Formulas_and_Stoichiometry/1.01%3A_Stoichiometry-_Chemical_Formulas_and_Equations/1.1.05%3A_Empirical_Formulas_from_Analysis]
• 12. Biomass Compositional Analysis Laboratory Procedures [https://www.nrel.gov/bioenergy/biomass-compositional-analysis]
• 13. Key ecological challenges for closed systems facilities [https://www.sciencedirect.com/science/article/abs/pii/S0273117713001725]
• 14. Integrated Pest Management Protocols for Space-Based ... [https://www.frontiersin.org/journals/astronomy-and-space-sciences/articles/10.3389/fspas.2021.759641/full]
• 15. Development of Nitrogen Recycling Strategies for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8548772/]
• 16. Mass Balances for a Biological Life Support System ... - PubMed [https://pubmed.ncbi.nlm.nih.gov/11537263/]
• 17. Using bioprocess stoichiometry to build a plant-wide mass ... [https://www.sciencedirect.com/science/article/pii/S0043135409000700]
• 18. Stoichiometry and Balancing Reactions [https://chem.libretexts.org/Bookshelves/Inorganic_Chemistry/Supplemental_Modules_and_Websites_(Inorganic_Chemistry)/Chemical_Reactions/Stoichiometry_and_Balancing_Reactions]
• 19. Flux Balance Analysis | Virginia - iGEM 2025 [https://2025.igem.wiki/virginia/fba]
• 20. The Biomass Objective Function - PMC - PubMed Central [https://pmc.ncbi.nlm.nih.gov/articles/PMC2912156/]
• 21. Flux Balance Analysis (FBA) [https://bio-protocol.org/exchange/minidetail?id=6366914&type=30]
• 22. Genome-scale metabolic model-guided systematic ... [https://www.nature.com/articles/s41540-025-00555-5]
• 23. FAQ: Metabolic Modeling [https://docs.kbase.us/workflows/metabolic-models/faq-metabolic-modeling]
• 24. Visualizing regulatory interactions in metabolic networks [https://bmcbiol.biomedcentral.com/articles/10.1186/1741-7007-5-46]
• 25. MINN: A metabolic-informed neural network for integrating ... [https://www.sciencedirect.com/science/article/pii/S2001037025003265]
• 26. Integration of proteomic data with genome‐scale metabolic ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11400636/]
• 27. Algorithms and tools for data-driven omics integration to ... [https://translational-medicine.biomedcentral.com/articles/10.1186/s12967-025-06446-x]
• 28. Multiomics Research: Principles and Challenges in ... [https://www.sciencedirect.com/science/article/pii/S2693125725000019]
• 29. Material balance and diet in bioregenerative life support systems [https://www.academia.edu/54511694/Material_balance_and_diet_in_bioregenerative_life_support_systems_Connection_with_coefficient_of_closure]
• 30. Omics approaches to study and model cell-cell interactions ... [https://www.frontiersin.org/journals/chemical-engineering/articles/10.3389/fceng.2025.1629455/full]
• 31. NEXT-FBA: A hybrid stoichiometric/data-driven approach to ... [https://www.sciencedirect.com/science/article/pii/S1096717625000461]
• 32. BLSS: a contribution to future life support [https://pubmed.ncbi.nlm.nih.gov/11537782/]
• 33. MELiSSA Pilot Plant [https://www.melissafoundation.org/page/melissa-pilot-plant]
• 34. Basic concepts and principles of stoichiometric modeling ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC4671265/]
• 35. Stoichiometric modelling of cell metabolism [https://www.sciencedirect.com/science/article/abs/pii/S1389172308700173]
• 36. Standardization of Human Metabolic Stoichiometric Models [https://www.frontiersin.org/journals/systems-biology/articles/10.3389/fsysb.2022.899980/full]
• 37. Recycling nutrients from organic waste for growing higher ... [https://www.sciencedirect.com/science/article/pii/S2214552423000639]
• 38. Metabolic Objectives and Trade-Offs: Inference and Applications [https://pmc.ncbi.nlm.nih.gov/articles/PMC11857637/]
• 39. Stoichiometric Modeling of Artificial String Chemistries ... [https://link.springer.com/article/10.1007/s00239-021-10018-0]
• 40. Elimination of Thermodynamically Infeasible Loops in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC3030201/]
• 41. ThermOptCobra: Thermodynamically optimal construction ... [https://www.sciencedirect.com/science/article/pii/S2589004225012660]
• 42. Thermodynamics-Based Metabolic Flux Analysis - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC1796839/]
• 43. What is flux balance analysis? - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3108565/]
• 44. Analyzing and Resolving Infeasibility in Flux Balance Analysis ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9317134/]
• 45. an efficient toolbox for sampling the loopless flux solution ... [https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-023-05616-2]
• 46. Mixed-Integer Optimization for Loopless Flux Distributions ... [https://arxiv.org/html/2502.00807v1]
• 47. 9. Loopless FBA — cobra 0.30.0 documentation [https://cobrapy.readthedocs.io/en/latest/loopless.html]
• 48. Mixed-Integer Optimization for Loopless Flux Distributions in ... [https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.SEA.2025.26]
• 49. https://public-pages-files-2025.frontiersin.org/journals/astronomy ... [https://public-pages-files-2025.frontiersin.org/journals/astronomy-and-space-sciences/articles/10.3389/fspas.2023.1198689/xml]
• 50. Optimization of the World Ocean Model of Biogeochemistry ... [https://bg.copernicus.org/articles/22/5349/2025/]
• 51. Coupling flux balance analysis with reactive transport ... [https://www.nature.com/articles/s41598-025-89997-9]
• 52. Multi-objective optimization can balance trade-offs among ... [https://www.nature.com/articles/s41598-022-15274-8]
• 53. Structure Data for Analysis [https://help.tableau.com/current/pro/desktop/en-us/data_structure_for_analysis.htm]
• 54. An efficient and flexible framework for inferring global ... [https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013427]
• 55. Global sensitivity analysis of integrated assessment models ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12411131/]
• 56. Publishing 13C metabolic flux analysis studies: A review and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5866723/]
• 57. High-resolution 13 metabolic flux analysis C [https://pubmed.ncbi.nlm.nih.gov/31471597/]
• 58. Validation-based model selection for 13 metabolic flux analysis with... C [https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1009999]
• 59. Model Validation and Selection in Metabolic Flux Analysis and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10055486/]
• 60. 13 C metabolic flux analysis [https://www.frontiersin.org/journals/molecular-neuroscience/articles/10.3389/fnmol.2022.883466/full]
• 61. OpenFLUX2: 13 - C software package adjusted for the... MFA modeling [https://microbialcellfactories.biomedcentral.com/articles/10.1186/s12934-014-0152-x]
• 62. Model Validation and Selection in Metabolic Flux Analysis and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10922127/]
• 63. Introduction to flux balance analysis (FBA) [https://www.protocols.io/view/introduction-to-flux-balance-analysis-fba-daz22f8e.html]
• 64. Optimization-based framework with flux balance analysis (FBA ... [https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1013635]
• 65. Predicting biological system objectives de novo from internal ... [https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-9-43]
• 66. Model Validation and Selection in Metabolic Flux Analysis ... [https://arxiv.org/abs/2303.12651]
• 67. Stoichiometric constraints and complete closure of long- ... [https://www.sciencedirect.com/science/article/abs/pii/S0273117704006210]
• 68. A Bayesian method for on-line evaluation of uncertainty in ... [https://www.sciencedirect.com/science/article/abs/pii/S026322412100436X]
• 69. Guidance for the Use of Bayesian Statistics in Medical ... [https://www.fda.gov/regulatory-information/search-fda-guidance-documents/guidance-use-bayesian-statistics-medical-device-clinical-trials]
• 70. A Tutorial on Modern Bayesian Methods in Clinical Trials [https://pmc.ncbi.nlm.nih.gov/articles/PMC10117244/]
• 71. BayFlux: A Bayesian method to quantify metabolic Fluxes ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10664898/]
• 72. Bayesian Analysis Reporting Guidelines [https://www.nature.com/articles/s41562-021-01177-7]
• 73. Bayesian approach to quantify parameter uncertainty and ... [https://link.springer.com/article/10.1007/s13762-013-0453-3]
• 74. Parallel labeling experiments and metabolic flux analysis [https://pmc.ncbi.nlm.nih.gov/articles/PMC6474253/]
• 75. Parallel labeling experiments and metabolic flux analysis [https://pubmed.ncbi.nlm.nih.gov/23246523/]
• 76. COMPLETE-MFA: complementary parallel labeling ... [https://pubmed.ncbi.nlm.nih.gov/24021936/]
• 77. Integrated 13C-metabolic flux analysis of 14 parallel labeling ... [https://pubmed.ncbi.nlm.nih.gov/25596508/]
• 78. Metabolic Flux Analysis—Linking Isotope Labeling and ... [https://www.mdpi.com/2218-1989/10/11/447]
• 79. Rational design of 13C-labeling experiments for metabolic flux ... [https://bmcsystbiol.biomedcentral.com/articles/10.1186/1752-0509-6-43]
• 80. Robustifying Experimental Tracer Design for13C-Metabolic ... [https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2021.685323/full]
• 81. Long-Duration Space Exploration: Stoichiometric Model of ... [https://www.ediweekly.com/long-duration-space-exploration-stoichiometric-model-of-a-fully-closed-bioregenerative-life-support-system-and-autonomous-robotics/]