Identification of CYP2B6 as a Novel Biomarker of HRD in Colon Adenocarcinoma through WGCNA and Machine Learning

The potential role of homologous recombination deficiency (HRD) in the diagnosis and treatment of colon adenocarcinoma (COAD) remains incompletely explored. Differential gene expression analysis was conducted using Limma to identify genes with altered expression levels. Key genes associated with HRD were identified through the integration of WGCNA and machine learning techniques. For the unsupervised grouping of samples, ConsensusClusterPlus was applied. To quantify gene expression and protein abundance in clinical tissues and cell lines, RT-qPCR and Western Blotting (WB) assays were performed, respectively. The "pRRophetic" package was employed to predict drug sensitivity profiles. Molecular docking simulations and optimal pose presentations were conducted by using CB-Dock2. Our comprehensive analysis of multiple COAD data sets, leveraging WGCNA and machine learning, unveiled five novel, previously unreported biomarkers of HRD: TNFRSF11A, SERPINA1, SPINK4, REG4, and CYP2B6. We devised an innovative HRD-linked molecular classification system and a predictive nomogram that accurately forecasts patient outcomes. Experimental validation substantiated the upregulation of CYP2B6 in COAD, enhancing proliferation and migration capabilities, and demonstrated a robust positive association with established HRD indicators RAD51 and γH2AX. Notably, CYP2B6 emerged as a promising predictor of PARP inhibitor (PARPi) sensitivity, offering potential therapeutic implications. In conclusion, our study, harnessing machine learning and experimental validation, has uncovered novel biomarkers of HRD and PARPi sensitivity, shedding light on potential avenues for tailored clinical treatment strategies in COAD, thereby advancing personalized medicine.