Identification of gastric mucosal carcinogenesis-related genes based on cross-sectional multi-stage samples:...

To identify the key genes differentially expressed across the gastritis-GC spectrum based on multi-stage gastric mucosal samples.

Five datasets associated with gastric mucosal carcinogenesis were obtained from the GEO database. DEGs between gastritis and GC were first determined. Four ML models, including LASSO, random Forest, SVM-RFE, and Boruta, were used to screen out DEGs. DCA and ROC analyses were performed to identify the highest discriminative potential model, followed by the identification of hub genes within the highest discriminative potential model. The expression pattern across the gastritis-GC spectrum and the potential pathways of hub genes were then explored. Their expression and prognosis differences in GC were then validated using TCGA data, followed by drug sensitivity assessment. RT-PCR, CCK-8, and transwell assays were used to detect the gene expression, cell viability, invasion, and migration.

Among 4 ML models, LASSO with 20 genes was the highest discriminative potential model to distinguish the gastritis from GC. After importance ranking on 20 genes, LILRA4 and IGF2BP3 were identified as the most important DEGs. Their expressions exhibited an upward trend within the gastritis-GC spectrum. A series of regression analyses, such as ordinal regression, trend regression, and GAM, all confirmed that their expression was significantly higher in GC compared to gastritis. Pathway analysis revealed that they were involved in metabolism and autophagy. LILRA4 was associated with GC status and negatively correlated with drug sensitivity. LILRA4 or IGF2BP3 overexpression promoted GC cell viability, invasion, and migration.

LILRA4 and IGF2BP3 may be associated genes differentially expressed across the gastritis-GC spectrum. They may have the potential to screen high-risk groups for GC in patients with gastritis.