The mechanisms underlying the resistance to immune checkpoint inhibitors (ICIs) therapy in metastatic urothelial carcinoma (mUC) patients are not clear. It is of great significance to discern mUC patients who could benefit from ICI therapy in clinical practice. In this study, we performed machine learning method and selected 10 prognostic genes for constructing the immunotherapy response nomogram for mUC patients. The calibration plot suggested that the nomogram had an optimal agreement with actual observations when predicting the 1- and 1.5-year survival probabilities. The prognostic nomogram had a favorable discrimination of overall survival of mUC patients, with area under the curve values of 0.815, 0.752, and 0.805 for ICI response (ICIR) prediction in the training cohort, testing cohort, and combined cohort, respectively. A further decision curve analysis showed that the prognostic nomogram was superior to either mutation burden or neoantigen burden for overall survival prediction when the threshold probability was >0.35. The immune infiltrate analysis indicated that the low ICIR-Score values in mUC patients were significantly related to CD8+ T cell infiltration and immune checkpoint-associated signatures. We also identified differentially mutated genes, which could act as driver genes and regulate the response to ICI therapy. In conclusion, we developed and validated an immunotherapy-responsive nomogram for mUC patients, which could be conveniently used for the estimate of ICI response and the prediction of overall survival probability for mUC patients.
Keywords: PD-L1; machine learning; metastatic urothelial carcinoma; nomogram; response.
Copyright © 2020 Chen, Zhang, Wang, Zhang, Wang and Zheng.