Abstract Antimetabolites, primarily studied in cancer, are novel drugs targeting metabolic networks by mimicking and inhibiting disease-causing metabolites, enabling poly-pharmacologic effects essential for complex diseases. Moreover, predicting patients likely to positively respond to antimetabolite drugs is necessary to simplify clinical applications. However, existing computational approaches for antimetabolite target discovery lack incorporation of disease-induced metabolic state perturbations, and their applicability beyond cancer remains unexplored. We introduce MATADOR (Metabolite-centric Analysis of TARgets for Drug ORientation), a computational workflow that integrates patient-derived omic data with metabolic networks to identify, evaluate and prioritize antimetabolite targets based on metabolic state transformation. Applying MATADOR to RNA-seq data from breast, colon, lung, and liver cancers, we achieved a 66±6% sensitivity in recapturing known antimetabolite targets, strongly supported thioredoxin as a pan-cancer target, and linked top-ranked targets to poor 5-year survival in breast and liver cancers. Extending beyond cancer, MATADOR nominated metabolites with proinflammatory effects as potential antimetabolite targets in Alzheimer's and Parkinson's diseases, aligning well with their known pathological mechanisms. Finally, applying MATADOR on personalized metabolic networks, machine learning models trained on metabolic gene expression demonstrated the ability to leverage gene expression to personalize antimetabolite targets. The proposed approach may expedite prioritization and personalization of antimetabolite targets during pre-clinical studies across diseases with systemic metabolic alterations.
Odongo et al. (2026) studied this question.