Development of an m6A subtype classifier to guide precision therapy for patients with bladder cancer

Purpose: Bladder cancer (BLCA) is a highly heterogeneous tumor. We aim to construct a classifier from the perspective of N6-methyladenosine methylation (m6A) to identify patients with different prognostic risks and treatment responsiveness for precision therapy. Methods: Data on gene expression profile, mutation, and clinical characteristics were mainly obtained from the TCGA-BLCA cohort. Unsupervised clustering was performed to construct m6A subtypes. The tumor microenvironment (TME) landscapes were explored by using ssGSEA, ESTIMATE, and MCPcounter algorithms. K-M survival curves and Cox regression analysis were used to demonstrate the significance of m6A subtypes in predicting prognosis. pRRophetic, oncoPredict, and TIDE algorithms were used to evaluate responsiveness to antitumor therapy. A classifier of m6a subtypes was finally developed based on random forest and artificial neural network (ANN). Results: The two m6A subtypes have significantly different m6A-related gene expression profiles and mutational landscapes. TME analysis showed a higher level of stromal and Inhibitory immune components in subtype B compared with subtype A. The m6A subtype is a clinically independent prognostic predictor of BLCA, subtype B has a poorer prognosis. Drug sensitivity analysis showed that subtype B has lower IC50 values and AUC values for cisplatin and docetaxel. Efficacy assessment showed significantly poorer radiotherapy efficacy and lower immunotherapy responsiveness in subtype B. We finally constructed an ANN classifier to accurately classify BLCA patients into two m6A subtypes. Conclusion: Our study developed a classifier for identifying subtypes with different m6A characteristics, and BLCA patients with different m6A subtypes have significantly different prognosis and responsiveness to antitumor therapy.