Machine learning-based identification of basement membrane-related signature to predict recurrence and immunotherapy benefit in bladder cancer.

The basement membrane (BM) plays a critical role in regulating bladder cancer (BC) progression. However, a BM-related signature for predicting BC recurrence has yet to be established. In this study, we developed a basement membrane-related signature (BRS) with 7 mRNAs using multiple machine learning algorithms based on data from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. The predictive performance of BRS for BC recurrence was evaluated via Kaplan-Meier survival and receiver operating characteristic (ROC) curves. A nomogram integrating BRS with clinical characteristics was developed and demonstrated superior clinical utility compared to individual parameters. Notably, immune infiltration analysis revealed distinct microenvironmental phenotypes between the two BRS groups: the low-BRS group was characterized by higher CD8+ T cell infiltration and reduced M0/M2 macrophages, indicating an immune-active microenvironment, whereas the high-BRS group exhibited lower CD8+ T cell infiltration and enrichment of M0/M2 macrophages, consistent with an immunosuppressive phenotype. Accordingly, the low-BRS group showed a more favorable predicted response to immunotherapy. Experimental data revealed that the upregulation of HIGD2A, one of the BRS gene, significantly suppressed BC cell proliferation, migration and invasion. It also downregulated MMP9 expression, suggesting a potential role in maintaining BM integrity. In conclusion, BRS serves as a novel and effective predictive tool for BC recurrence, and therapeutic targeting of HIGD2A may offer a viable strategy to mitigate the risk of disease progression.Clinical trial registration: Not applicable.