Prognostic development and validation of a prediction model based on major histocompatibility complex-related differentially expressed genes in stomach adenocarcinoma.

Abstract

Background: Stomach adenocarcinoma (STAD) is a common malignant tumor with high morbidity and mortality. Major histocompatibility complex (MHC) is an important component of the immune system responsible for antigen presentation. However, no studies have yet reported on the relationship between major histocompatibility complex-related differentially expressed genes (MHCRDEGs) and the survival prognosis of STAD. The aim of this study is to explore the relationship between MHCRDEGs and survival prognosis in STAD patients.

Methods: Using The Cancer Genome Atlas (TCGA) database, we screened for differentially expressed MHCRDEGs, and a survival prognosis model was constructed based on these genes. We generated training and validation samples from the TCGA and Gene Expression Omnibus (GEO) datasets to enhance the robustness of our findings. The predictive effects of the model were assessed using Kaplan-Meier (KM) survival curve analysis, receiver operating characteristic (ROC) curve analysis, calibration analysis and decision curve analysis (DCA), with statistical significance reported as P values. The differences in the expression of key MHCRDEGs between different subgroups of TCGA and GEO databases were analyzed. Finally, a multifactorial survival prognostic model was constructed by combining MHC score (MHCs), and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was used to verify the expression of key genes.

Results: We identified five key MHCRDEGs: MKI67, MYB, SERPINE1, TRIM31, and HAVCR1. In the first prognostic model, the KM curves demonstrated a highly statistically significant difference in predicting overall survival (OS) in patients (P<0.001). The ROC curves indicated that the model showed relatively low accuracy in predicting 1-year [area under curve (AUC) =0.616], 3-year (AUC =0.644), and 5-year (AUC =0.619) occurrence. Furthermore, calibration analysis and DCA suggested that the model's predictions of OS were consistent with the actual patient survival, with the 5-year prognostic model exhibiting the best clinical utility. In the TCGA and GEO datasets, most of the key genes showed significant expression differences between the STAD/GEO and normal groups (P<0.001). Finally, the predictive model constructed by combining MHCs with clinicopathological staging demonstrated good predictive accuracy with optimal clinical utility at 5 years, with specific accuracy metrics provided as part of our results, and validated their expression via qRT-PCR in cell lines (MKI67: P=0.01, MYB: P=0.02, SERPINE1: P=0.02, TRIM31: P=0.02, HAVCR1: P<0.0001).

Conclusions: In this study, the expression and distribution of MHCRDEGs in STAD were analyzed by various methods, and a clinical prediction model of STAD was constructed using MHCRDEGs. The validity of this model confirms the feasibility of MHCRDEGs as prognostic markers for STAD, elucidating their potential clinical implications in guiding treatment strategies for this disease.