Document resource
Background To characterize the landscape of genetically influenced metabotypes (GIMs) underlying gastric cancer (GC) development, and to evaluate their utility for precision prevention.Methods We developed a causal stable learning (CSL) framework to infer 249 genetically determined metabolomic traits by integrating genome-wide variants and metabolomic data from 436,480 participants across five cohorts: the UK Biobank (UKBB) Discovery (n=145,857) and Test (n=284,694) cohorts, and cohorts from the Upper Gastrointestinal Cancer Early Detection program (UGCED, n=370), Shandong Intervention Trial (SIT, n=2755), and Mass Intervention Trial in Linqu, Shandong province (MITS, n=2804). GIMs were established through integrative trans-omics analyses, with additional metabolomic profiling, single-cell sequencing, and functional experiments to elucidate downstream effects. The utility of GIMs to enhance GC prevention was evaluated in two randomized trials.Results The CSL models identified 118 putative causal genomic loci for 228 metabolomic traits, achieving high prediction consistency among the UKBB and other external cohorts ( IDDF2026-ABS-0384 Figure 1. Causal stable learning framework characterized the basis of GIMs for gastric cancer). The genetically predicted traits yielded 1604 significant trait-metabolite correlation pairs (P-value<0.001), suggesting close relations between genetically regulated metabolic susceptibility and gastric carcinogenesis. Overall, the genetically predicted traits displayed directionally consistent associations with GC risk across the independent cohorts. We further established an online database to characterize GIM landscapes for GC and gastric lesion progression with 71 traits and 77 putative causal loci, indicating highly consistent cross-ancestry regulatory effects. Among these, four predicted very-low-density lipoprotein triglyceride traits were positively associated with progression from intestinal metaplasia to gastric neoplasia in SIT and UGCED subjects. Blood metabolomics analyses further showed that long-chain triglyceride levels were associated with increased lesion progression risk (OR=1.56, 95% CI:1.30-1.86). At single-cell resolution, these precited traits increased along the transition from enterocyte-like cells to precancerous lesion cells and malignant cells, and were positively correlated with OLFM4 expression (Spearman’s r=0.19-0.53). Combined in-silico prediction and functional evidence suggested PLTP as a key metabolic effector gene in gastric carcinogenesis (IDDF2026-ABS-0384 Figure 2. GIMs revealed the effects of germline variations on gastric cancer susceptibility). In randomized trials, GIMs improved GC risk stratification and facilitated the characterization of treatment heterogeneity in H. pylori eradication for GC prevention (IDDF2026-ABS-0384 Figure 3. GIMs enabled refined risk stratification and precision prevention of gastric cancer).Conclusions Genetically encoded metabolic individuality contributes to GC susceptibility, providing translational potential for enabling precision prevention of GC.Abstract IDDF2026-ABS-0384 Figure 1Abstract IDDF2026-ABS-0384 Figure 2Abstract IDDF2026-ABS-0384 Figure 3