A degradome-related signature for predicting the prognosis and immunotherapy benefit in stomach adenocarcinoma based on machine learning procedure

Abstract

Stomach adenocarcinoma (STAD) is one of the subtype of gastric cancer with high invasiveness, extreme heterogeneity, high morbidity, and high mortality. The degradome is the most abundant class of cellular enzymes that play an essential role in regulating cellular activity and carcinogenesis. An integrative machine learning procedure including 10 methods was performed to develop a prognostic degradome-based prognostic signature (DPS) in TCGA, GSE15459, GSE26253, and GSE62254 datasets. Investigations of the DPS concerning immune infiltration, immunotherapy benefits, and drug priority were orchestrated. The DPS developed by Enet [alpha = 0.3] method was regarded as the optimal prognostic model. The DPS had a stable and powerful performance in predicting the clinical outcome of STAD and served as an independent risk factor in training and testing cohorts. The C-index of DPS was higher than that of age, sex, and clinical stage. STAD patients with low DPS scores had a higher abundance of B cells, CD8+ T cells, higher cytolytic scores, and T cell co-stimulation scores. Moreover, low DPS score indicated a lower tumor immune dysfunction and exclusion score, lower T cell dysfunction and exclusion score, higher PD1&CTLA4 immunophenoscore, and higher tumor mutation burden score in STAD, demonstrating a better immunotherapy response. STAD patients with a high DPS score had a lower IC50 value of common chemotherapy and targeted therapy regimens (Cisplatin, Docetaxel, Gefitinib, etc). Our study developed an optimal DPS for STAD. The DPS could predict the prognosis, risk stratification and guide treatment for STAD patients.