A transcriptomic study for identifying cardia- and non-cardia-specific gastric cancer prognostic factors using...

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Gastric cancer (GC) is heterogeneous, with distinct features between cardia and non-cardia subtypes.
• Existing prognostic models often lack site-specificity, limiting their predictive accuracy.
• There is a need for tailored prognostic tools for different GC locations.

Purpose of the Study:

• To identify site-specific transcriptomic prognostic biomarkers for cardia and non-cardia gastric cancer.
• To develop and validate novel prognostic models for these distinct GC subtypes.
• To enhance the accuracy of survival prediction for personalized gastric cancer treatment.

Main Methods:

• Utilized the Cancer Genome Atlas (TCGA) database for transcriptomic data analysis.
• Employed genetic algorithm-based support vector machine (GA-SVM) and Cox regression (GA-Cox) for biomarker identification.
• Evaluated model performance using time-dependent ROC curve analysis (AUC) and concordance index (C-index).

Main Results:

• Identified 10 prognostic biomarkers for cardia cancer and 13 for non-cardia cancer.
• Site-specific models demonstrated significantly improved 5-year survival prediction compared to traditional models (cardia: AUC 0.899, non-cardia: AUC 0.994).
• Combined nomograms showed superior predictive performance (C-index cardia=0.816, non-cardia=0.812).

Conclusions:

• Developed accurate, site-specific prognostic models for cardia and non-cardia gastric cancer.
• These models offer improved survival prediction, supporting individualized therapy decisions for GC patients.
• A user-friendly online tool (GC-SMS) was created for accessible application of these findings.