Integrating multi-omics and machine learning to decipher the molecular pathways of bisphenol a-associated lactylation-related genes driving bladder cancer

In this study, we systematically investigated bladder cancer–related gene signatures using a toxicogenomics-informed framework, with particular attention to genes associated with lactylation-related pathways. Multi-omics data from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) were integrated, and Weighted Gene Co-expression Network Analysis (WGCNA), a toxicology database, and lactylation-related gene sets were combined for intersection screening. Machine learning algorithms, including LASSO, SVM, and random forest, were then applied to identify key genes. Four prioritized BPA–lactylation-associated candidate genes—ENO1, WBP11, GTF2F1, and SPR—were ultimately identified and showed consistent associations with metabolic, immune, and transcription-related features. Multi-level validation, including immune infiltration analysis, single-cell transcriptome localization, proteomic validation, and molecular docking and kinetic simulation, supported the structural plausibility of BPA–protein interactions at the molecular level. This study proposes a toxicogenomics-informed, hypothesis-generating framework that prioritizes candidate genes and pathways potentially linking BPA-related signatures with lactylation-associated processes in bladder cancer.