Clinical-genomic stratification of treatment biomarkers and KMT2C/D-linked MSI heterogeneity in Chinese gastric cancer: a retrospective cohort study

Biomarker-guided therapy has reshaped gastric cancer treatment, but the clinical-genomic features underlying biomarker-defined treatment access remain incompletely characterized in Chinese patients. We aimed to define clinically relevant genomic stratification across treatment-related biomarkers, metastatic phenotypes, and KMT2C/D-associated molecular subgroups in advanced gastric cancer. We retrospectively analyzed paired tumor-normal targeted sequencing data from 563 patients with stage III/IV gastric cancer or gastroesophageal junction adenocarcinoma (QZ563 cohort), including 391 patients with detailed clinicopathologic, biomarker, treatment, and survival data (Clinical391 cohort). Somatic alterations, tumor mutational burden, microsatellite instability, metastatic phenotype, treatment-related biomarkers, and outcomes were integrated with TCGA-STAD transcriptomic data and DepMap cancer-dependency resources. TP53 (70.8%), ARID1A, CDH1, and PIK3CA were among the most frequently mutated genes. Treatment biomarkers and metastatic phenotype captured additional clinically relevant heterogeneity, including a small, molecularly heterogeneous biomarker-negative subgroup warranting further validation. KMT2C/D loss-of-function (LoF) alterations, present in 6.4% of the QZ563 cohort, were strongly enriched in MSI-H tumors (48.3% vs. 2.7%, P < 0.001) and associated with higher TMB within both MSS and MSI-H strata, but not with a survival benefit; this molecular pattern was also observed in TCGA-STAD. DepMap analyses further nominated WRN as a candidate dependency in KMT2C/D-LoF MSI models. This clinical-genomic analysis characterizes biomarker-related treatment-access patterns and metastatic phenotype-associated heterogeneity in advanced Chinese gastric cancer. KMT2C/D-LoF further stratifies MSI/TMB-high tumors and nominates WRN dependency as a hypothesis-generating vulnerability requiring functional validation. This study was approved by the Ethics Committee of Peking University Cancer Hospital (approval no. 2019KT37; February 11th, 2019).