Abstract

Hypoxia, a hallmark of solid tumors, drives malignant progression and represents a major therapeutic challenge. Metabolic reprogramming induced by hypoxia creates unique metabolic vulnerabilities that can be exploited therapeutically. Here, we systematically compared the metabolic network differences between hypoxic and normoxic cells, and developed DepFormer, a transformer-based deep learning model, to nominate hypoxia-dependent metabolic genes in tumor cells. Oxidative phosphorylation was identified as the most significantly hypoxia-dependent metabolic pathway, and FLAD1 was predicted to be one of the key hypoxia-dependent metabolic genes. FLAD1 locus is amplified, and FLAD1 expression is upregulated across various tumor types, especially in hypoxic tumors. FLAD1 depletion disrupts activity of mitochondrial complex II, causing succinate/fumarate imbalance, which in turn prevents cancer cells from adapting to hypoxia. We further identified a drug-like inhibitor of FLAD1, which selectively inhibits growth of hypoxic tumor cells. Our study establishes DepFormer as an effective framework for predicting state-specific metabolic dependencies and reveals FLAD1 as a metabolic vulnerability and an innovative therapeutic target for hypoxic tumors.