Identification of endoplasmic reticulum stress-related biomarkers in idiopathic pulmonary fibrosis based on machine learning

Abstract

Abstract Background Endoplasmic reticulum stress (ERS) is critical in the development and progression of idiopathic pulmonary fibrosis (IPF). The aim of this study was to explore ERS-related biomarkers in IPF using a bioinformatics approach and to further investigate their relationship with immune cells. Three IPF datasets (GSE10667, GSE24206 as the training set and GSE53845 as the validation set) were obtained based on the Gene Expression Omnibus (GEO). In the training set, ERS-related differentially expressed genes(DEGs) between IPF and normal tissues were screened and subjected to GO and KEGG enrichment analysis. Key ERS-related DEGs were further screened using Weighted Gene Co-expression Network Analysis (WGCNA) and three machine learning algorithms (Least Absolute Shrinkage and Selection Operator (LASSO), Random Forest (RF), and Support Vector Machine - Recursive Feature Elimination with Local Feature Selection (SVM-RFE)) and validated in a validation set. Then the CIBERSORT method was used to calculate the immune cell infiltration abundance and investigate the relationship between immune cells and key markers. Results we obtained 65 ERS-related DEGs from the training set, and 2 key ERS-related DEGs (COMP, GPX8) were screened by WGCNA and machine learning and validated in the validation set. COMP and GPX8 showed high diagnostic value (AUC > 0.8). The results of immune cell infiltration studies showed substantial associations between these two key markers and T-cell CD8, neutrophils, monocytes, macrophage M2 and plasma cells. Conclusion COMP and GPX8 are valuable potential biomarkers for IPF and provide a basis for future studies on the early diagnosis and treatment of IPF.