Development and validation of a ligand-receptor pairs signature to predict outcome and provide a therapeutic strategy in gastric cancer.

BACKGROUND: An important factor in tumor development and progression is the tumor microenvironment (TME), which is heterogeneous. Previous studies have mainly investigated the expression profile and prognostic values of genes in gastric cancer (GC) at the cell population level but neglected the interactions and heterogeneity between cells.

METHODS: The pattern of ligand-receptor (LR) interactions was delineated in a scRNA-seq dataset containing 44953 cells from nine GC patients and fourth bulk RNA-seq datasets including data from 1159 GC patients. We then constructed a LR.Score scoring model to comprehensively evaluate the influence of LR-pairs on the TME, overall survival, and immunotherapy response in GC patients from several cohorts.

RESULTS: Cell communication network among 13 cell types was constructed based on the LR-pairs. We proposed a new molecular subtyping model for GC based on the LR-pairs and revealed the differences in prognosis, pathophysiologic features, mutation characteristics, function enrichment, and immunological characteristics among the three subtypes. Finally, an LR. Score model based on the LR-pairs was developed and validated in several datasets.

CONCLUSIONS: Based on the selected LR-pairs, we successfully constructed a novel prediction model and observed its well performance on molecular subtyping, targets and pathways screening, prognosis judging, and immunotherapy response predicting.