Objective:

To assess whether genetically predicted plasma protein levels influence gastric adenocarcinoma risk and to identify druggable candidates using a proteome-wide Mendelian randomization framework.

Approach:

• Mendelian Randomization: Integrated deCODE plasma protein quantitative trait loci (pQTLs) with gastric cancer GWAS in a proteome-wide two-sample MR framework.
• Protein Interaction Analysis: Used protein–protein interaction topology to prioritize eight hub proteins for further evaluation.
• Artificial Neural Network Classifier: Developed an eight-gene ANN classifier to distinguish between gastric cancer and non-tumor tissue.
• Expression Mapping: Conducted single-cell and spatial transcriptomics, along with multiplex immunofluorescence, to map hub-gene expression.
• Target Focus: Focused on TGFB3 as a target, combining in silico ligand screening with in vitro perturbation of the TGFB3–PI3K survival axis.

Key Findings:

• Identified 29 circulating proteins with putative causal effects on gastric cancer, including eight central hub proteins.
• The eight-gene ANN classifier showed robust discrimination between gastric cancer and non-tumor tissue.
• TGFB3 was enriched at tumor–stroma interfaces and associated with poor survival.
• Proflavine hemisulfate attenuated TGFB3-driven proliferation and migration in AGS cells.

Interpretation:

The study identifies TGFB3 as a genetically supported candidate involved in gastric cancer.

Limitations:

• Mendelian randomization does not provide direct proof of protein-level functional causality within tumor tissue.
• Findings may be influenced by the biological heterogeneity of gastric cancer.

Conclusion:

The study identifies TGFB3 as a tractable target for therapeutic strategies in gastric adenocarcinoma.