Integrative multi-omics and machine learning identify a robust signature for discriminating prognosis and therapeutic targets in bladder cancer

Background: Bladder cancer (BLCA) is a common malignant tumor whose pathogenesis has not yet been fully elucidated. This study analyzed prognostic genes in BLCA by integrating transcriptomics and proteomics data, and established prognostic models, aiming to offer novel insights for BLCA therapy. Methods: Transcriptomic, proteomic, and protein acetylation sequencing were conducted on six BLCA tumor tissues and six paraneoplastic tissue samples. Furthermore, data from TCGA-BLCA, GSE13507, and single-cell RNA sequencing (scRNA-seq) datasets were integrated. Initially, differential expression analysis identified candidate genes regulated by acetylation. These genes were further refined by intersecting with scRNA-DEG obtained from the scRNA-seq dataset, resulting in the identification of key genes. Subsequently, consistency clustering analysis was performed based on these key genes. Prognostic models were then developed utilizing Cox regression analysis and least absolute shrinkage and selection operator (LASSO) Cox regression. Independent prognostic factors were determined through independent prognostic analysis, followed by the establishment of a nomogram model. Additionally, gene set enrichment analysis (GSEA), immune cell infiltration analysis, mutation analysis, and drug sensitivity analysis were conducted between the two risk groups to elucidate underlying mechanisms. Results: A total of 15 key genes were obtained by crossing 284 candidate genes with 510 scRNA-DEGs. Patients in the TCGA-BLCA dataset were categorized into two subtypes based on the 15 key genes. Next, a risk model was developed using five prognostic genes (CTSE, XAGE2, MAP1A, CASQ2, and FXYD6), and a nomogram model was developed using age, pathologic T, pathologic N, and risk score. A total of 1089 GO entries and 49 KEGG pathways, including cytokine-cytokine receptor interactions, ECM receptor interactions, etc., were involved in all genes in both risk groups. The immunization score, matrix score, and ESTIMATE score were significantly higher in the low-risk group than in the high-risk group. Conclusion: CTSE, XAGE2, MAP1A, CASQ2 and FXYD6 were selected as prognostic genes in BLCA, risk model and nomogram model predicting the prognosis of BLCA patients were constructed. These were helpful for prognostic assessment of BLCA.