Background
Immune checkpoint inhibitors (ICIs) have significantly transformed the treatment of gastroesophageal cancer (GEC). However, the lack of reliable prognostic biomarkers hinders the ability to predict patient response to ICI therapy.
Methods
In this study, we engineered and validated a genomic mutation signature (GMS) utilizing an innovative artificial intelligence (AI) algorithm to forecast ICI therapy outcomes in GEC patients. We further explored immune profiles across subtypes through comprehensive multiomics analysis. Our investigation of drug sensitivity data from the Genomics of Drug Sensitivity in Cancer (GDSC) database led to the identification of trametinib as a potential therapeutic agent. We subsequently evaluated trametinib’s efficacy in AGS and MKN45 cell lines using Cell Counting Kit-8 (CCK8) assays and clonogenic experiments.
Results
We developed a GMS by integrating 297 algorithms, enabling autonomous prognosis prediction for GEC patients. The GMS demonstrated consistent performance across three public cohorts, exhibiting high sensitivity and specificity for overall survival (OS) at 6, 12, and 18 months, as shown by Receiver Operator Characteristic Curve (ROC) analysis. Notably, the GMS surpassed traditional clinical and molecular features, including tumor mutational burden (TMB), programmed death-ligand 1 (PD-L1) expression, and microsatellite instability (MSI), in predictive accuracy. Low-risk samples exhibited elevated levels of cytolytic immune cells and heightened immunogenic potential compared to high-risk samples. Our investigation identified trametinib as a potential therapeutic agent. An inverse correlation was observed between GMS and trametinib IC50. Moreover, the high-risk-derived AGS cell line showed increased sensitivity to trametinib compared to the low-risk-derived MKN45 cell line.
Conclusion
The GMS utilized in this study successfully demonstrated the ability to reliably predict the survival advantage for patients with GECs undergoing ICI therapy.
Supplementary Information
The online version contains supplementary material available at 10.1007/s12672-024-01400-7.