FluxAtlas is a pan-cancer metabolic flux atlas that identifies tumor-selective dependencies and prognostic biomarkers

Metabolic reprogramming is a hallmark of cancer, yet dynamic metabolic flux has been difficult to study systematically. Here, we present FluxAtlas, a pan-cancer atlas of metabolic flux generated from genome-scale metabolic modeling of over 10,000 tumors across 28 The Cancer Genome Atlas (TCGA) cancer types. By integrating enzyme constraints and nutrient diffusion limits, we reveal conserved and tissue-specific metabolic rewiring, including alterations in bile acid recycling, urea metabolism, and amino acid biosynthesis. We identify a bile acid-associated program that remodels glutathione homeostasis and drives gastrointestinal-specific lipid metabolism. Comparisons with matched normal models uncover tumor-selective metabolic dependencies, such as increased reliance of renal cancers on de novo purine synthesis. Under nutrient limitation, modeling predicts convergence on glutamine-dependent aspartate synthesis while preserving tissue-specific metabolic states. Machine learning models based on fluxomics predict patient survival and highlight biotin uptake as a prognostic biomarker. FluxAtlas defines the functional metabolic landscape of human cancer and is accessible at https://software.icr.ac.uk/app/flux-atlas.