Integrated Multi-omics Analysis and Machine Learning Identify Hub Genes and Potential Mechanisms of Resistance to Immunotherapy in Gastric Cancer

Background: Patients with gastric cancer respond poorly to immunotherapy. There are still unknowns about the biomarkers associated with immunotherapy sensitivity and their underlying molecular mechanisms.

Methods: Gene expression data for gastric cancer were gathered from TCGA and GEO databases. DEGs associated with immunotherapy response came from ICBatlas. KEGG and GO analyses investigated pathways. Hub genes identification employed multiple machine algorithms. Associations between hub genes and signaling pathways, disease genes, immune cell infiltration, drug sensitivity, and prognostic predictions were explored via multi-omics analysis. Hub gene expression was validated through HPA and CCLE. Multiple algorithms pinpointed Cancer-Associated Fibroblasts genes (CAFs), with ten machine-learning methods generating CAFs scores for prognosis. Model gene expression was validated at the single-cell level using the TISCH database.

Results: We identified 201 upregulated and 935 downregulated DEGs. Three hub genes, namely CDH6, EGFLAM, and RASGRF2, were unveiled. These genes are implicated in diverse disease-related signaling pathways. Additionally, they exhibited significant correlations with disease-associated gene expression, immune cell infiltration, and drug sensitivity. Exploration of the HPA and CCLE databases exposed substantial expression variations across patients and cell lines for these genes. Subsequently, we identified CAFs-associated genes and established a robust prognostic model. The analysis in the TISCH database showed that the genes in this model were highly expressed in CAFs.

Conclusions: The results unveil an association between CDH6, EGFLAM, and RASGRF2 and the immunotherapeutic response in gastric cancer. These genes hold potential as predictive biomarkers for gastric cancer immunotherapy resistance and prognostic assessment.