An H. pylori-related transcriptional gene panel for prognostic stratification in gastric cancer: an exploratory transcriptomic and immunohistochemical study

Helicobacter pylori infection is a major risk factor for gastric cancer (GC), but the prognostic relevance of host transcriptional programs related to H. pylori remains incompletely defined. This study developed an H. pylori -related transcriptional gene panel for exploratory prognostic stratification and examined its associations with immune-related molecular features. Single-cell RNA sequencing (GSE163558), bulk transcriptomic profiles from TCGA-STAD, and two independent GEO cohorts (GSE62254 and GSE84433) were integrated. An H. pylori -related pathway activity score derived from 71 curated host genes was used as a transcriptional surrogate rather than a measure of bacterial infection. The original cell-level score comparison and WGCNA were used for exploratory candidate screening. A six-gene panel was constructed in TCGA-STAD; the same six genes were evaluated in the GEO cohorts on their platform-specific processed expression scales using cohort-specific risk estimation and median stratification. Immune infiltration, TIDE outputs, immune checkpoint markers, mutation profiles, and predicted drug sensitivity were compared between risk groups. IHC compared four available proteins in 18 GC patients with documented H. pylori status (positive, n = 9; negative, n = 9); this analysis did not constitute validation of the complete six-gene prognostic model. Nine major cell types were identified by single-cell analysis, and a six-gene panel (CXCR4, TNFAIP2, TIMP1, TAP1, RGS2, and DUSP1) was constructed. In the TCGA development cohort, the high-risk group had worse overall survival (HR = 2.41, 95% CI 1.78–3.27, log-rank P < 0.0001; C-index = 0.65, 95% CI 0.60–0.70; 1/2/3-year AUC = 0.69/0.66/0.62). The same gene panel also stratified survival in two independent cohorts, although those analyses were cohort-specific rather than validation of a frozen cross-platform equation. The age-plus-RiskScore nomogram showed moderate apparent discrimination (C-index = 0.68, 95% CI 0.63–0.73). Computational analyses suggested a less immune-favorable profile in the high-risk group. Three of four IHC-evaluated proteins (DUSP1, RGS2, and TIMP1) showed higher expression in H. pylori -positive tumors after FDR correction. The six-gene panel showed exploratory prognostic stratification across three retrospective cohorts. Because the public transcriptomic datasets lacked patient-level H. pylori infection status and the cross-platform evaluations were cohort-specific, these findings do not establish an infection-specific or directly transportable prognostic model. IHC supports an association between documented infection status and three selected proteins, but not validation of the complete six-gene signature. Larger prospective studies are required.