Metabolic Regulation Classification Scheme

metabolic-regulation-classification-scheme

Summary

A concordance-based classification method that assigns metabolic reactions into four regulatory categories (combined transcriptional and metabolic, metabolic-only, transcriptional-only, or unclassified) by measuring Cohen's kappa agreement between RAS-vs-FFD and RPS-vs-FFD variation signs across pairwise cell-line comparisons. This skill discriminates whether differences in metabolic flux originate from transcriptional control of enzyme abundance or from metabolic control via substrate availability.

When to use

When you have computed Reaction Activity Scores (RAS) from transcriptomics and GPR rules, Reaction Presence Scores (RPS) from RAS normalized flux predictions, and Flux Fold-change Distributions (FFD) from metabolomic data and mass-action constraints across multiple sample pairs, and you need to assign regulatory mechanism(s) to each reaction to guide therapeutic targeting or metabolic engineering design.

When NOT to use

• When reactions lack GPR associations—these reactions cannot be included in RPS-RAS concordance analysis and will have missing RPSvsRAS values.
• When a single substrate for a reaction is missing from metabolomics measurements—the entire reaction must be omitted from FFD-based classification.
• When you have only cross-sectional (single timepoint) transcriptomics or metabolomics data and cannot form meaningful pairwise comparisons across ≥2 biological conditions.

Inputs

• Reaction Activity Scores (RAS) computed from transcriptomics data and GPR rules for all reactions and sample pairs
• Reaction Presence Scores (RPS) from normalized RAS-derived flux constraints within constraint-based models
• Flux Fold-change Distributions (FFD) sampled from flux feasible regions using Mann–Whitney U test on metabolomics-constrained models
• Irreversible metabolic model in SBML or MAT format (reactions and GPRs split)
• Cell-line pair identifiers (10 pairwise comparisons minimum)

Outputs

• Classification table with columns: reaction identifier, RASvsFFD Cohen's kappa, RPSvsRAS Cohen's kappa, empirical p-value, FDR-adjusted p-value, assigned regulatory class (combined/metabolic-only/transcriptional-only/unclassified)
• Concordance analysis dataset with Cohen coefficients and Pearson correlations for each pairwise comparison

How to apply

For each of the 10 pairwise cell-line comparisons, compute the sign of variation (up +1, down −1, no-change 0) for RAS using t-test (p<0.05) with fold-change threshold ≥20%, and for RPS and FFD using Mann–Whitney U test (p<0.05). Then calculate two Cohen's kappa coefficients for each reaction: RASvsFFD and RPSvsRAS, which quantify concordance of variation signs across all 10 pairwise comparisons. Classify reactions based on kappa sign patterns: positive both scores → combined transcriptional and metabolic control; positive RPSvsFFD and negative RPSvsRAS → metabolic-only control; negative both scores with high RASvsFFD → transcriptional-only control; positive RPSvsRAS but negative RPSvsFFD → unclassified/other. Apply Benjamini–Hochberg FDR correction to empirical p-values derived from randomized RPS sampling, setting significance threshold at FDR<5%. Compile final classification table with reaction identifiers, both kappa scores, adjusted p-values, and assigned regulatory class.

Related tools

• COBRApy (Constraint-based modeling library used to construct cell-line-specific models and sample flux distributions with optGpSampler for FFD computation) — https://github.com/opencobra/cobrapy
• constraint-based stoichiometric metabolic models (Scaffold for integrating transcriptomics (via RAS) and metabolomics (via FFD) to predict reaction-level flux changes)
• mannWhitneyUTest.py (Computes Mann–Whitney U test to determine FFD and RPS variation signs (up/down/no-change) across pairwise comparisons) — https://github.com/qLSLab/integrate
• rasTtest.py (Computes t-test on RAS scores to determine transcriptional variation signs across pairwise comparisons) — https://github.com/qLSLab/integrate
• concordanceAnalysis.py (Orchestrates Cohen's kappa concordance calculation, FDR correction, and final regulatory classification of reactions) — https://github.com/qLSLab/integrate

Examples

python pipeline/concordanceAnalysis.py --valLog 1.2 --weight linear --resultsMetabolomicFile resultsMetabolomic --metabolic_model ENGRO2_irrev.xml --lcellLines MCF102A SKBR3 MCF7 MDAMB231 MDAMB361 --meansFile medie_Met.csv

Evaluation signals

• All 10 pairwise cell-line comparisons yield valid variation sign matrices (up/down/no-change) for RAS, RPS, and FFD with no missing values for included reactions.
• Cohen's kappa values fall within the expected range [−1, +1] and satisfy FDR<5% threshold after Benjamini–Hochberg correction; kappa>0.41 typically indicates moderate to substantial concordance.
• Reactions are distributed across all four regulatory classes with counts that reflect known biology (e.g., more combined regulation in central metabolism, more transcriptional-only in anabolic pathways).
• When visualizing a sample reaction (e.g., ACONT as in the README), variation signs are consistent across both RASvsFFD and RPSvsRAS comparisons, matching the assigned regulatory class.
• Final classification table contains no reactions with both RASvsFFD and RPSvsRAS = NaN unless they lack GPR associations or have missing metabolomics substrates—these omissions are documented.

Limitations

• Reactions without GPR associations cannot be classified because RPS cannot be computed; these must be pre-filtered before concordance analysis.
• When any single reaction substrate is missing from metabolomics measurements, the entire reaction is omitted from FFD computation and concordance analysis, reducing coverage of metabolic network.
• Direct metabolic flux determination through labeled substrates is not employed; FFD is inferred via mass-action law on metabolomics data and flux sampling, which may not capture non-steady-state dynamics or allosteric effects not encoded in the model.
• Classification robustness depends on statistical power of pairwise comparisons; sparse or noisy transcriptomics or metabolomics data may increase false-positive or false-negative regulatory assignments.
• The scheme assumes that variation sign concordance reflects causal regulation; bidirectional feedback loops or indirect metabolic control may be misclassified as transcriptional-only or metabolic-only.

Evidence

• [other] Reactions are classified by measuring Cohen's kappa concordance between RAS-vs-FFD and RPS-vs-FFD variation signs: positive concordance for both indicates combined regulation, positive RPSvsFFD with negative RASvsFFD indicates metabolic control only, and positive values for both RASvsFFD and RASvsRPS indicates transcriptional and metabolic regulation.: "Reactions are classified by measuring Cohen's kappa concordance between RAS-vs-FFD and RPS-vs-FFD variation signs: positive concordance for both indicates combined regulation, positive RPSvsFFD with"
• [other] For each reaction, calculate two Cohen's kappa coefficients: RASvsFFD and RPSvsRAS, quantifying concordance of variation signs across the 10 pairwise comparisons.: "For each reaction, calculate two Cohen's kappa coefficients: RASvsFFD and RPSvsRAS, quantifying concordance of variation signs across the 10 pairwise comparisons."
• [other] Apply Benjamini–Hochberg FDR correction to empirical p-values derived from randomized RPS sampling to set significance threshold at FDR<5%.: "Apply Benjamini–Hochberg FDR correction to empirical p-values derived from randomized RPS sampling to set significance threshold at FDR<5%."
• [intro] The intersection of the two output datasets discriminates fluxes regulated at the metabolic and/or gene expression level: "The intersection of the two output datasets discriminates fluxes regulated at the metabolic and/or gene expression level"
• [abstract] We discriminate fluxes regulated at the metabolic and/or transcriptional level. This information is valuable to better inform targeted action planning in different fields, including personalized prescriptions in multifactorial diseases, such as cancer: "knowing the regulatory level at which a given metabolic reaction is controlled will be valuable to inform targeted, truly personalized therapies in cancer"
• [results] Missing RPSvsRAS values occur when a reaction is not associated with a GPR: "Missing RPSvsRAS values occur when a reaction is not associated with a GPR"
• [results] If one single reaction substrate is missing from the metabolomics measurements, the reaction is omitted from the dataset: "If one single reaction substrate is missing from the metabolomics measurements, the reaction is omitted from the dataset"
• [readme] Aim: create the results (dataset and figures) for concordance analysis. A dataset of concordance analysis (Cohen coefficient and pearson correlation) of RPS vs RAS, RPS vs FFD, RPS vs FFD, the pvalues and the adjusted pvalues: "A dataset of concordance analysis (Cohen coefficient and pearson correlation) of RPS vs RAS, RPS vs FFD, the pvalues and the adjusted pvalues"