Reference-guided computational framework identifies microenvironment metabolic subtypes and targets using pan-cancer single-cell datasets

作者：Ke Tang, Ya Han, Dongqing Sun, Xin Dong, Tong Han, Hailin Wei, Wenwen Shao, Junjie Hu, Zhaoyang Liu, Lele Zhang, Taiwen Li, Peng Zhang, Qiu Wu, Chenfei Wang · 发表于：Genome Medicine · 年份：2025 · DOI：10.1186/s13073-025-01572-z · 被引用次数：1 · 研究领域：Single-cell and spatial transcriptomics、Ferroptosis and cancer prognosis、Immune cells in cancer

BACKGROUND: Metabolic reprogramming is a hallmark of cancer; however, the mechanisms driving metabolic heterogeneity across diverse cell types in the tumor microenvironment remain poorly understood. Most existing methods predict metabolic states at the pathway level but rarely map reaction-level alterations to their upstream regulators, thereby constraining both interpretability and translational relevance. METHODS: We developed MetroSCREEN, a reference-guided computational framework that infers reaction-level metabolic flux propensity and nominates upstream regulators from bulk and single-cell transcriptomes. MetroSCREEN uses a fast enrichment-based procedure to quantify reaction-level metabolic activity. To characterize metabolic regulons, it integrates intrinsic gene-regulatory signals with extrinsic cell-cell interaction cues, then applies a robust multi-evidence ranking scheme to combine these information sources, and finally employs a constraint-based causal discovery module to infer regulatory directionality. RESULTS: MetroSCREEN accurately predicts reaction-level metabolic activities and their upstream regulators, as demonstrated using paired transcriptomic-metabolomic datasets from the cancer cell lines. We further validated predicted regulators with in-house single-cell CRISPR screens in PC9 cells targeting metabolic regulators. Applying MetroSCREEN to a pan-cancer single-cell atlas of more than 700,000 fibroblasts and myeloid cells across 36 cancer types, we identi...