CytoGem-XAI:A Hypergraph Neural Network Framework for Genome-Scale Metabolic Modeling and Interpretable Analysis
Abstract
Genome-scale metabolic models are essential for understanding cellular metabolism, yet existing deep learning approaches remain black boxes, and traditional flux balance analysis (FBA) cannot provide sample-specific predictions. To our knowledge, CytoGem-XAI is the first framework to combine hypergraph neural network representation with interpretable, FBA-parallel analysis and sample-specific metabolic characterization. Built upon hypergraph representations where reactions are encoded as hyperedges connecting their participating metabolites, CytoGem-XAI introduces three analysis modules: perturbation-based carbon source importance ranking, hard intervention reaction bottleneck identification, and pathway-level topological attribution. Beyond prediction, CytoGem-XAI uniquely enables condition-dependent carbon source essentiality and reaction bottlenecks that vary with genetic background--capabilities absent from both traditional FBA and existing deep learning methods. Trained on 17,400 E. coli growth conditions using 10-fold cross-validation, our framework achieves R2 = 0.862, substantially outperforming AMN (R2 = 0.81, +6.4%), FBA (R2 = 0.62, +39%), and gradient boosting baselines (R2 = 0.71, +21%). Biological validation confirms that CytoGem-XAI identifies known essential carbon sources (e.g., alanine, malate) and rate-limiting enzymes (e.g., TCA cycle), while also revealing N-acetylmuramate--a peptidoglycan precursor--as a previously underappreciated essential nutrient.
Full-text reasoning
From the deep-tier full-text analysis of this preprint.
Reasoning review — 7 key claims, 1 well-supported, 6 with gaps
- supported CytoGem-XAI achieves high predictive accuracy (R^2=0.862) and substantially outperforms AMN, FBA, and gradient boosting.
- partial CytoGem-XAI uniquely combines hypergraph representation, FBA-parallel interpretable analysis, and sample-specific metabolic characterization. gap: The claim of uniqueness and being the "first" is an assertion based on a summarized literature review, not a detailed comparative analysis presented in the text.
- partial CytoGem-XAI identifies known essential carbon sources (e.g., alanine, malate) and rate-limiting enzymes (e.g., TCA cycle). gap: The text states that biological validation confirms this but does not present the validation methodology or results to support the claim.
- partial CytoGem-XAI provides interpretable insights through perturbation, intervention, and topological weighting modules. gap: The text describes the design of interpretable modules but does not demonstrate the interpretability or the direct parallelism of its insights with classical FBA workflows.
- partial CytoGem-XAI enables sample-specific metabolic characterization, revealing condition-dependent essentiality unavailable in FBA. gap: The text claims this capability but does not demonstrate how it reveals condition-dependent essentiality or provide examples of sample-specific characterization.
- partial Hypergraph neural networks can learn biologically meaningful representations, recapitulating decades of knowledge. gap: The text asserts that the pathway attribution quantitatively recapitulates knowledge but does not present the quantitative data or analysis supporting this claim.
- unsupported CytoGem-XAI reveals N-acetylmuramate as a previously underappreciated essential nutrient. gap: The text states this as a discovery but provides no evidence or methodology for how CytoGem-XAI revealed it or why it is considered essential and underappreciated.
Claims and gaps are read from the full text by a language model, shown for transparency; they do not affect ranking or selection.
Lifecycle
- biorxiv v1 2026-06-07 source ↗
Discussion
No qualifying discussion yet.