A Critical Re-evaluation of "Nutrient competition predicts gut microbiome restructuring under drug perturbations" by Shi et al., Cell 2025; doi:10.1016/j.cell.2025.10.038

This work provides a comprehensive, critical, and figure-resolved re-evaluation of the study by Shi et al. (Cell, 2025), which proposed that nutrient competition is the primary mechanistic driver of gut microbiome restructuring under drug perturbations. While the original study integrates metabolic modeling, in vitro assays, ecological simulations, and gnotobiotic mouse experiments, much of its mechanistic interpretation rests on inferred nutrient profiles, computationally constructed nutrient fingerprints, and ecological models that presuppose nutrient-based competition as the foundational organizing principle.

In this commentary, we assess the entire study figure-by-figure, including all Main Figures, Extended Data Figures (1–15), and Supplementary Figures, offering an in-depth critique of the experimental design, computational methodology, statistical interpretation, and biological plausibility of the conclusions. Particular emphasis is placed on four recurrent methodological limitations:

1. circularity in nutrient preference inference, where genome-based annotations are treated as ground truth despite high uncertainty;
2. model overfitting and confounder leakage, where nutrient indices may inadvertently encode drug physicochemical features or baseline microbiome structure;
3. lack of direct nutrient or metabolomic measurements, preventing empirical validation of nutrient-shift hypotheses; and
4. insufficient causal testing in vivo, as gnotobiotic experiments do not manipulate nutrient pools or measure host-mediated responses known to modulate microbial communities.

The commentary highlights how monoculture growth assays using supraphysiological drug concentrations cannot isolate nutrient competition effects; how ecological models restrict interaction structures in ways that enforce nutrient competition by design; and how nutrient fingerprints remain speculative in the absence of direct metabolomic confirmation. Extended Data and Supplementary Figures reveal additional concerns, including parameter sensitivity, inconsistent nutrient annotations, and underpowered validation cohorts.

Collectively, the analysis argues that while nutrient competition may contribute to drug-induced community shifts, the current evidence does not justify treating it as the dominant or mechanistic driver. Instead, drug toxicity, stress responses, immune modulation, bile acid dynamics, mucus glycan remodeling, and strain-level heterogeneity represent equally plausible—and often untested—mechanisms. By delineating these limitations, this work provides a rigorous, constructive, and transparent framework for refining mechanistic studies of drug–microbiome interactions and for guiding future experimental designs capable of distinguishing nutrient-driven ecology from alternative biological processes.