Computational integration of single-cell transcriptomics and functional dependency data reveals metabolic-regulatory vulnerabilities of drug-tolerant persister cells in EGFR-mutant non-small-cell lung cancer

Drug-tolerant persister (DTP) cells survive targeted therapy through reversible, non-genetic adaptations and can seed eventual relapse, yet how their metabolic reprogramming unfolds over time—and what regulatory logic governs it—remains poorly understood. We addressed this gap computationally, analyzing time-resolved single-cell RNA-seq data from PC9 EGFR-mutant lung cancer cells treated with erlotinib across six time points (3,046 cells after quality control). Rather than a single metabolic switch, we found a staged transition: an acute shift toward oxidative phosphorylation within the first 24 hours, a subsequent collapse of lipogenic programs linked to inferred SREBF1/2 transcription factor repression, and a late consolidation of aldehyde-detoxification and ferroptosis-defense pathways. These changes persist after cell cycle regression, ruling out proliferative arrest as the sole driver. Upstream, a regulatory network anchored by NFE2L2 (NRF2) connects 12 hub transcription factors to the metabolic effector genes. When we cross-referenced these candidates against DepMap CRISPR screens stratified by a DTP-like expression signature, GPX4 emerged as preferentially essential in DTP-signature-high cell lines (P=4.7×10−4), whereas the thioredoxin reductase TXNRD1 was broadly essential regardless of DTP status. An independent PC9 dataset treated with gefitinib and osimertinib reproduced the core metabolic changes. Because this is a purely computational study, every target and therapeutic hypothesis including the prioritization of auranofin for combination with EGFR inhibitors requires experimental validation. The data and framework are offered as a resource for designing such experiments.