An Integrative Genome-Scale Metabolic Modeling and Machine Learning Framework for Predicting and Optimizing Single-Cell Protein Production in Saccharomyces cerevisiae
Authors: Neha K. Nair, Aaron D'Souza
Organizations: Department of Physics and Chemistry, National Institute of Technology, Warangal, Telangana, India · Department of Electronics and Communication Engineering, National Institute of Technology, Warangal, Telangana, India
Saccharomyces cerevisiae is increasingly recognised as a key source for single-cell protein (SCP) production, a rising solution to global protein-supply challenges. This study presents a computational framework combining the Yeast9 genome-scale metabolic model (GEM) with machine learning and optimisation to predict and enhance biomass flux for SCP yield. The Yeast9 GEM, comprising 4,131 reactions, 2,806 metabolites, and 1,161 genes, was simulated using flux balance analysis (FBA) across 2,000 Latin Hypercube-sampled flux profiles. Random Forest and XGBoost regressors achieved R2 values of 0.9999760 and 0.9997702, respectively. A variational autoencoder (VAE) identified four metabolic clusters with mean biomass fluxes of 0.472, 0.493, 0.527, and 0.505 gDW/hr. SHAP-based feature attribution identified twenty key reactions in glycolysis, the TCA cycle, and amino-acid biosynthesis; 18/20 (90%) were confirmed essential by in silico knockout. Bayesian optimisation produced a 12.13-fold improvement in biomass flux (0.0858 to 1.041 gDW/hr) at glucose = -20.0, oxygen = -20.0, and ammonium = -8.9 mmol/gDW/hr. A generative adversarial network (GAN) generated novel flux configurations (variance = 0.124); stoichiometric feasibility verification returned 0/100 feasible profiles due to incomplete generator convergence, reported as a limitation. Pareto front analysis identified an optimal SCP operating point at 0.0858 gDW/hr biomass flux with amino-acid biosynthesis score of 1000.029 mmol/gDW/hr.
Figures & tables
Parameter
Value / Description
Model
Consensus Yeast GEM (Yeast9 v9.0.2)
Reactions
4,131
Metabolites
2,806
Genes
1,161
Biomass objective reaction
r_2111
Optimal biomass flux (default)
0.08584 gDW ⋅ hr -1
Table 1: Yeast9 GEM characteristics and FBA simulation outputs.
Cluster
Mean Biomass Flux (gDW ⋅ hr -1 )
Rank
0
0.472428
Lowest
1
0.493054
3rd
2
0.526680
Highest
3
0.504600
2nd
Table 2: Mean biomass flux per metabolic cluster.
Figure 1: VAE Training Loss
Figure 2: Random Forest: True vs Predicted Biomass
Figure 3: Feature Importance
Figure 4: Shap-Guided in Silico Overexpression
Figure 5: Bayesian Optimization Convergence
Figure 6: Oxygen Sensitivity Analysis
Biomass lower bound (gDW ⋅ hr -1 )
AA biosynthesis flux (mmol ⋅ gDW -1 ⋅ hr -1 )
Feasibility
Note
0.000
1000.029
Feasible
Knee point
0.025
1000.029
Feasible
0.051
1000.029
Feasible
0.076
1000.029
Feasible
Upper limit (default glucose)
0.101
—
Infeasible
Carbon-limited at r_1714 = − 1.0
0.127–0.989
—
Infeasible (36/40 pts)
Requires glucose <− 10 mmol ⋅ gDW -1 ⋅ hr -1
Table 3: Pareto front results: biomass vs. amino-acid biosynthesis flux.
Figure 7: Pareto Front Analysis
Feature
Previous research
Current study
SCP optimisation
Most workflows lack computational optimisation pipelines, SHAP interpretation, flux analysis, or genome-scale metabolic modelling.
Integrates generative AI, SHAP analysis, machine learning, and optimisation for SCP production.
GEM with ML
Primarily focused on growth prediction rather than SCP optimisation; low interpretability and limited generative discovery.
Integrates RF, XGBoost, FFNN, SHAP, nutrient optimisation, and SCP production with Yeast9 GEM simulations.
Generative AI
Lacked SCP focus, explainability, or integrated strain-engineering workflows; primarily applied to kinetic or cancer metabolism.
Explores metabolic states using VAE and GAN frameworks for SCP-relevant metabolic engineering.
Bayesian optimisation
Insufficient integration with explainable AI, flux-level analysis, or SCP optimisation.
Combines SCP optimisation with GEM metabolic simulations via Bayesian optimisation.
SHAP analysis
Limited to broad reaction interpretation with minimal mechanistic resolution.
Uses SHAP to determine the influence of metabolic reactions on biomass and amino-acid biosynthesis.
Multi-objective optimisation
Pareto optimisation rarely used in SCP-specific metabolic engineering.
Applies Pareto-front optimisation between amino-acid biosynthesis and biomass to improve SCP nutritional value.
Table 4: Comparison of the current study with previous research across key methodological features.
Department of Computing, Imperial College London, United Kingdom · BASF SE, Ludwigshafen, Germany · School of Mathematical Sciences, Lancaster University, United Kingdom +1