AI and Cancer Research ── Bladder ── 2026-11-08
Bladder cancer: treatment, trials and AI
- Abstract
INTRODUCTION: Bladder cancer is the most prevalent malignancy of the urinary tract. This study aimed to identify candidate biomarkers for bladder cancer through an integrated multi-omics analysis. METHODS: Single-cell RNA sequencing (scRNA-seq), bulk RNA sequencing (RNA-seq), and proteomic datasets were analyzed to characterize cellular populations and identify differentially expressed genes (DEGs) and proteins (DEPs). Candidate genes were prioritized using four machinelearning algorithms, with model interpretation performed using SHapley Additive exPlanations (SHAP). Gene set enrichment analysis (GSEA) was conducted to explore associated pathways, whereas Enrichr/DSigDB enrichment and molecular docking were performed to investigate potential therapeutic compounds and compound-gene interactions. A five-gene SVM model was externally validated in the independent bulk transcriptomic cohort GSE13507. The expression levels of the candidate genes were further evaluated in bladder cancer cell lines. RESULTS: Fifteen major cell populations were identified. Exploratory analyses highlighted endothelial perivascular stromal cells and inflammatory cancer-associated fibroblasts as cell populations of interest, with predicted intercellular communication patterns suggesting potential roles in the bladder cancer microenvironment. Moreover, 7 genes/proteins showed concordant upregulation and 122 downregulation across all three datasets, and these candidates were subsequently subjected to machine- learning analysis. CFD, WNT2B, MAPK13, ABLIM1, and PRKAR2B were prioritized as five candidate genes for bladder cancer, and their expression patterns and potentially associated pathways were further evaluated in bladder cancer samples and cell lines. Among the four evaluated algorithms, SVM demonstrated the highest reported internal holdout-set AUC of 1.000 (95% CI: 1.000-1.000), and was selected for SHAP interpretation. In the independent GSE13507 cohort, the five-gene SVM model achieved an AUC of 0.815 (95% CI: 0.752-0.877). Molecular docking suggested potential interaction poses between Rifaximin and CFD, and between MAPK13 and NG- 25/XMD13-2. DISCUSSION: The integration of multi-omics data with an interpretable SHAP-based model provides an exploratory framework for prioritizing candidate bladder cancer biomarkers and generating hypotheses for subsequent biological and clinical validation. CONCLUSION: This multi-omics study identified candidate biomarkers and exploratory compoundgene associations for bladder cancer. These findings require validation in additional independent multicenter cohorts and functional studies before clinical application.
Journal IF-equivalent: 4.6 (OpenAlex 2-year mean citedness, value as of 2026-11-06, retrieved 2026-11-07; not the official Clarivate IF)Reference: Guan H, Liu X, Hong T, Li H, Liu G, Zhang Y, et al. Integrative Multi-Omics Analysis with Interpretable Machine Learning Prioritizes Candidate Biomarkers for Bladder Cancer. CGT. 2026 Oct 27;26. doi:10.2174/0115665232538656261006070001.Checked: Abstract only