Multiple programmed cell death patterns and immune landscapes in bladder cancer: Evidence based on machine learning and multi‐cohorts

Abstract Background Bladder cancer (BLCA) is the most prevalent malignant neoplasm of the urinary tract, and ranks seventh as the most frequent systemic neoplasm in males. Dysregulation of programmed cell death (PCD) has been implicated in various stages of cancer progression, including tumorigenesis, invasion, and metastasis. However, the correlation between multiple PCD modes and BLCA is lacking. Thus, a risk prediction model was built based on 12 models of PCD to predict prognosis and immunotherapy response in patients with BLCA. Methods The RNA sequencing transcriptome data of BLCA were collected from the Cancer Genome Atlas Program (TCGA) and GEO datasets. Univariate Cox and LASSO regression analyzes were performed to identify PCD‐related genes (PCDRGs) significant for prognosis. Multivariate Cox regression analysis was used to develop a prognostic model for PCD. Survival analysis and chi‐squared test were employed to analyze the survival variations between different risk groups. Univariate and multivariate Cox analyses were performed to evaluate the model as an independent prognostic predictor. A nomogram was formulated using both clinical data and the model to predict the survival rates of BLCA patients. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes were performed to analyze and elucidate the molecular mechanisms and pathways operating within different risk score groups. Furthermore, the immune landscape was investigated and the efficacy of various anti‐tumor drugs was evaluated for BLCA. Finally, consensus clustering analysis was adopted to explore the association between different PCD clusters and clinical characteristics. Results Assessment of the public datasets and multivariate Cox analysis yielded 1254 PCDRGs, of which 10 PCDRGs for BLCA were identified. Based on the PCDRGs, a prognostic model was built for BLCA patient prognosis. Compared with the low‐risk group, the high‐risk group had a poorer prognosis. The model predicted area under the curve (AUC) values of 0.751, 0.753, and 0.763, respectively, for 1‐, 3‐, and 5‐year survival of BLCA patients. The nomogram further demonstrated the credibility of the prognosis model. The low‐risk group patients exhibited lower TIDE scores and higher TMB scores, implying better response of the low‐risk group to immunotherapy. The consensus clustering analysis indicated that compared with PCD cluster A, PCD cluster B was significantly more expressed in PCDRGs, suggesting a closer relation of PCD cluster B to PCDRGs. Patients in PCD cluster B had lower risk scores. Conclusion To summarize, the effects of 12 PCD patterns on BLCA were synthesized and the correlation between PCD and BLCA was explored. These findings provide new and convincing evidence for individualized treatment of BLCA, and help guide the treatment strategy and improve the prognosis of BLCA patients.