Machine Learning-Driven Identification and In Vitro Validation of the APOBEC3B-ANLN Regulatory Axis in Adrenocortical Carcinoma.

BACKGROUND: Adrenocortical carcinoma (ACC) is a rare, aggressive malignancy with limited diagnostic and therapeutic options. APOBEC3B (A3B) has emerged as a mutational driver in several cancers, but its downstream mechanisms remain unclear. We aim to utilize bioinformatics methods, such as machine learning, to reveal the mechanism of A3B in ACC and verify and explore it in depth in vitro. METHODS: Through the comprehensive analysis of 311 samples, including differential expression analysis and weighted gene co-expression network analysis (WGCNA), we use the hub genes extracted from the key modules as the background for 113 machine learning methods. Genes with potential associations were evaluated using feature importance and SHAP analysis techniques, and in vitro studies included qRT-PCR, Western blotting, siRNA-mediated knockdown, overexpression rescue, scratch assays, and Transwell migration assays to assess effects on gene expression and cell motility. RESULTS: Random Forest was selected as the optimal model and identified nine gene features centered on A3B and ANLN (AUC  = 0.996). Knockout of the A3B gene significantly reduced the mRNA and protein levels of ANLN (p  < 0.001). ANLN overexpression rescued the outcome. Compared with knockout alone, the cell migration distance and the number of migrating cells were restored (p  < 0.001). CONCLUSIONS: Our comprehensive omics and experimental methods have revealed the A3B-ANLN axis as a key mechanism of ACC and also provided predictive models and potential targets for the early diagnosis and therapeutic intervention of ACC.