# Integrative Mendelian Randomization and Pathomics Analysis Using Expression Quantitative Trait Loci and Genome‐Wide Association Study Data Identifies Mismatch Repair Genes as Prognostic Biomarkers in Gastric Adenocarcinoma

## TL;DR

This study shows that mismatch repair genes, especially MSH2, are important for predicting outcomes in stomach cancer and can be estimated using histopathology images.

## Contribution

A novel image-based model was developed to predict MSH2 expression, integrating genomic and histopathologic data for gastric cancer prognosis.

## Key findings

- MLH1 and PMS2 were identified as risk genes, while MSH2 showed a protective effect in gastric cancer.
- The pathomics model predicted MSH2 expression with an AUC of 0.811 using histopathology images.
- High-survival-probability groups had elevated tumor mutational burden and TP53 mutations.

## Abstract

Mismatch repair (MMR) genes are implicated in stomach adenocarcinoma (STAD). This study assessed their causal role in STAD, prognostic value, and developed a histopathology‐based model to predict MutS homolog 2 (MSH2) expression.

Using data from the IEU OpenGWAS database, five Mendelian randomization (MR) models evaluated causal links between MMR genes and gastric cancer (GC). Prognostic relevance was assessed via survival analysis. A random forest model using TCGA hematoxylin and eosin–stained images was trained to predict MSH2 expression. Biological insights were explored via pathomics score, gene set enrichment analysis (GSEA), immune infiltration, and tumor mutational burden (TMB).

MR analysis identified MLH1 and PMS2 as risk genes, while MSH2 had a protective effect. Cox regression confirmed MSH2 as an independent protective factor (HR = 0.690, 95% CI: 0.487–0.977, p < 0.05). The pathomics model predicted MSH2 expression with an AUC of 0.811. A comparison of high‐ and low‐survival‐probability (SP) groups showed differentially expressed genes, including SFRP4. The high‐SP group had elevated TMB and TP53 mutation frequency.

MMR genes, especially MSH2, are critical in STAD development and prognosis. The image‐based model effectively predicts MSH2 expression, supporting the integration of genomic and histopathologic data for personalized GC care.