Explore the scholarly record.

Abstract

Abstract Background: GERD is a common chronic digestive disorder, yet its diagnosis remains challenging due to the lack of objective molecular markers are lacking. Pyroptosis may contribute to esophageal mucosal injury, but a systematic search for pyroptosis-related diagnostic genes in GERD has not been performed. In this study, we used integrated bioinformatics, machine learning, and experimental validation to identify pyroptosis-associated diagnostic biomarkers for GERD and examine their relationship with the esophageal immune microenvironment. Methods: Esophageal transcriptome data from patients with GERD were obtained from the GEO database. We intersected differentially expressed genes, WGCNA module genes, and a curated set of pyroptosis-related genes to obtain candidate genes. We constructed a diagnostic model from 113 combinations of machine learning algorithms and validated it in an independent dataset using an artificial neural network. Immune cell infiltration was estimated with CIBERSORT, and the key bioinformatics results were further checked in a rat model of GERD. Results: The machine learning pipeline identified five pyroptosis-related genes with diagnostic value: GABARAPL1, CD55, CA9, SLC4A2, and LCN2. Among these, GABARAPL1 and CD55 performed best individually, with AUCs of 0.876 and 0.862. When the two genes were fed into an artificial neural network, the AUC reached 0.919 in the training cohort and 0.900 in the test cohort. Immune profiling showed that GERD esophageal mucosa contained more resting/activated dendritic cells, M2 macrophages, and activated NK cells, but fewer CD8⁺ T cells and γδ T cells. The animal experiments confirmed that GABARAPL1 was downregulated and CD55 was upregulated in GERD tissues. Conclusion: These observations point to GABARAPL1 and CD55 as potential pyroptosis-related diagnostic biomarkers for GERD. They may also influence disease pathogenesis through immunomodulatory mechanisms.