Molecular Markers and Diagnostic Model Specific for Barrett's Esophagus.

Abstract

Biomarkers involved in the progression of Barrett's esophagus (BE) have not been extensively studied. We aimed to identify novel molecular markers for the early diagnosis of BE. The expression profiles of GSE100843 including BE segment and normal squamous mucosa samples before and after vitamin D3 supplementation were downloaded from Gene Expression Omnibus. Differentially expressed genes (DEGs) were identified using the limma package. Principal component analysis was performed using Minitab, and DEGs in the top three principal components were clustered into different gene sets by the mclust package. Pathways and functions enriched by these gene sets were evaluated by deregulation score analysis. Key genes associated with BE were identified by coexpression analysis and a genetic algorithm. Using the xgboost package, an XGBoost classifier specific for BE was further constructed based on the key genes. A total of 2598 DEGs were identified, which were further clustered into nine gene sets. According to the deregulation scores of pathways and functions enriched by these gene sets, nine functional and pathway terms were significantly deregulated in BE. Among the DEGs, CREB3L1, HNF1B, and IL35 were genes with high fitness levels and connectivity degrees, predicting that they were key genes associated with BE. The XGBoost classifier constructed using the key genes was efficient and robust in BE prediction. The accuracies for prediction were 93% and 87% for training and validation datasets, respectively. Key genes may serve as novel biomarkers of BE, and the XGBoost classifier may contribute to the diagnosis of BE in future clinical practice.