Identification of CLDN3 as a Novel Biomarker for Gastric Precancerous Lesions via Comprehensive Bioinformatics and Machine Learning Analysis

Abstract

Abstract Gastric cancer (GC) poses a serious threat to human health, and early detection and treatment of Low-grade intraepithelial neoplasia (LGIN) is crucial in preventing its progression to GC. Therefore, it's important to identify decisive diagnostic biomarkers of LGIN for effective treatment selection. To achieve this, we obtained two datasets from the GEO database(GSE130823, GSE55696), using the GSE55696 as a validation set. Our analysis, using various bioinformatics strategies and machine-learning algorithms, led to the identification of 328 differentially expressed genes (DEGs) primarily involved in processes such as organic anion transport, regulation of blood pressure, and cell-cell junction assembly. Our results suggest that CLDN3 is a potentially valuable diagnostic biomarker for LGIN, as determined through algorithms such as LASSO regression, SVM-RFE, and WGCNA, and confirmed through receiver operating characteristic curves and CIBESORT analysis. Our findings indicate that CLDN3 is a reliable diagnostic biomarker for gastric cancer precursors, as confirmed by validation sets and immunohistochemical staining images in HPA. Additionally, we established a TF-CLDN3-miRNA regulatory network, which provides insights into the regulatory mechanisms of CLDN3 in the precancerous process of gastric cancer. We also found that immune cell infiltration may play a crucial role in the development of LGIN, and that CLDN3 has potential regulatory mechanisms in this process.