Metabolomics in Cancer Drug-Response Prediction and Target Discovery: From Experimental Models to Clinical Translation

Drug response in cancer often remains heterogeneous even among tumors sharing the same genomic driver, indicating that genotype alone does not fully capture the metabolic state underlying therapeutic sensitivity. Metabolomics provides a window into this state by measuring the biochemical consequences of oncogenic signaling, nutrient availability, microenvironmental stress, and drug perturbation, although metabolite abundance does not directly quantify pathway flux. In this narrative review, we examine how cancer pharmacometabolomics is evolving from descriptive biomarker profiling toward drug-response prediction and metabolic target discovery. We organize the evidence across three translational settings: high-throughput cancer models for identifying response-associated metabolic states, patient-derived and in vivo systems for testing metabolic dependencies and resistance mechanisms, and clinical cohorts for treatment stratification and longitudinal monitoring. We further examine isotope tracing, spatial and single-cell metabolomics, multi-omics modeling, and causal inference as complementary approaches for interpreting metabolic changes. We propose five non-cumulative evidence checkpoints (E1–E5) spanning treatment-linked association or prediction, internal evaluation, independent corroboration, source/pathway or target-specific perturbation evidence, and prospective clinical utility. Convergent evidence from predictive modeling, tracer-informed pathway analysis, cross-model corroboration, and genetic or pharmacological perturbation can strengthen target hypotheses. The proposed checkpoints provide a qualitative appraisal framework rather than a formally validated score, and clinical signatures still require independent evaluation before routine use.