Unraveling the influence of metabolic signatures on immune dynamics for predicting immunotherapy response and survival in cancer

Abstract

Abstract Metabolic reprogramming in cancer significantly impacts immune responses within the tumor microenvironment, but its influence on cancer immunotherapy effectiveness remains uncertain. This study aims to elucidate the prognostic significance of metabolic genes in cancer immunotherapy through a comprehensive analytical approach. Utilizing data from the IMvigor210 trial (n = 348) and validated by retrospective datasets, we performed patient clustering using non‐negative matrix factorization based on metabolism‐related genes. A metabiotic score was developed using a “DeepSurv” neural network to assess correlations with overall survival (OS), progression‐free survival, and immunotherapy response. Validation of the metabolic score and key genes was achieved via comparative gene expression analysis using qPCR. Our analysis identified four distinct metabolic classes with significant variations in OS. Notably, the metabolism‐inactive and hypoxia‐low class demonstrated the most pronounced benefit in terms of OS. The metabolic score predicted immunotherapeutic benefits with high accuracy (AUC: 0.93 at 12 months). SETD3 emerged as a crucial gene, showing strong correlations with improved OS outcomes. This study underscores the importance of metabolic profiling in predicting cancer immunotherapy success. Specifically, patients classified as metabolism‐inactive and hypoxia‐low appear to derive substantial benefits. SETD3 is established as a promising prognostic marker, linking metabolic activity with patient outcomes, advocating for the integration of metabolic profiling into immunotherapy strategies to enhance treatment precision and efficacy.

Keywords

• cancer
• deep learning model
• immunotherapy
• metabolism
• tumor microenvironment