Integrative transcriptomic and machine-learning analysis of the STAT1–GOLM1 axis in lung adenocarcinoma and small-molecule inhibitor discovery

Tumor invasion and distant metastasis remain major causes of treatment failure in lung adenocarcinoma (LUAD), thereby elucidating the molecular mechanisms is critical for precise treatment of LUAD. Using bulk RNA-Seq and single cell transcriptome data, combining with multi-machine learning methods, we identified GOLM1 as a core gene associated with poor prognosis in LUAD, which is highly expressed in malignant epithelial cells characterized by increased extracellular matrix (ECM)-integrin signaling, glycosylation remodelling and invasion-related transcriptional programs. Mechanistically, we validated that STAT1 is an upstream transcription activator of GOLM1 that forms a STAT1-GOLM1 axis that triggers LUAD invasion. Candidate compounds were prioritized by computational molecular docking against STAT1 and subsequently evaluated in A549 cells. Experimental validation revealed that Compound 1 exhibited the strongest inhibitory effect on LUAD cells by regulating PI3K/AKT signaling, cell-cycle and apoptosis. Collectively, our results elucidated the significant role of STAT1-GOLM1 axis in LUAD and provided a promising AI-directed inhibitor for LUAD.