An AI-Driven Multi-Omics Framework Identifies CASP8 as a Clinically Actionable Pyroptosis Biomarker in Bladder Cancer

Despite rapid advances in multi-omics technologies, translating candidate biomarkers into clinical practice for bladder cancer remains challenging due to the difficulty of linking complex genomic instability to interpretable biological processes. To address this, we developed an AI-driven multi-omics discovery framework integrating single-cell RNA sequencing, multi-cohort transcriptomics, and machine learning-based genomic inference. By analyzing chromosomal aneuploidy and copy number variations at single-cell resolution, we identified malignant cell populations and constructed a consensus pyroptosis scoring system, followed by machine learning-assisted biomarker screening and experimental validation. Our results reveal that while global pyroptosis activity is elevated in the bladder cancer microenvironment, malignant cells with high genomic instability exhibit significant pyroptosis suppression. Through this pipeline, CASP8 was identified as a key clinically relevant biomarker; its low expression correlates strongly with increased tumor mutation burden, frequent driver gene alterations (including TP53 and RB1), and poor survival outcomes. Functional assays further confirmed that CASP8 loss promotes malignant phenotypes and alters cell death programs. Ultimately, this study establishes a next-generation framework for biomarker translation, highlighting CASP8 as a clinically actionable link between genomic instability and pyroptosis dysregulation, and demonstrating the power of AI-integrated strategies in accelerating bladder cancer research from bench to bedside.