Modeling microbiome modulation of tumor metabolic networks to predict synergistic therapies.

Differences in microbiome composition profoundly influence drug response, yet methods to model the metabolic impact of microbes on host cells and therapeutics remain limited. We present a microbiome-aware computational framework combining machine learning and genome-scale metabolic models to predict combination therapies for colorectal cancer (CRC) in the presence of Fusobacterium nucleatum (Fn) and other pathogenic, probiotic, and commensal microbes. The model learned predictive metabolic flux signatures from 6,514 drug combination profiles in CRC cell lines and predicted synergistic drug combinations across both microbe-free and microbe-associated contexts. Model performance was supported through prospective comparison with newly reported drug combinations, in vitro drug synergy assays, microbiome co-culture experiments, and targeted metabolic perturbations of predicted pathway dependencies. Pharmacological perturbations in asymmetric co-cultures revealed phosphoinositol metabolism and cysteine transport as key determinants of Fn-dependent drug synergy. Together, this work introduces a scalable strategy for discovering microbiome-dependent combination therapies, including chemotherapies, immunotherapy, and probiotics.