Renal fibrosis is a physiological and pathological characteristic of chronic kidney disease (CKD) to end-stage renal disease. Since renal biopsy is the gold standard for evaluating renal fibrosis, there is an urgent need for additional non-invasive diagnostic biomarkers.
We used R package “limma” to screen out differently expressed genes (DEGs) based on Epithelial-mesenchymal transformation (EMT), and carried out the protein interaction network and GO, KEGG enrichment analysis of DEGs. Secondly, the least absolute shrinkage and selection operator (LASSO), random forest tree (RF), and support vector machine-recursive feature elimination (SVM-RFE) algorithms were used to identify candidate diagnostic genes. ROC curves were plotted to evaluate the clinical diagnostic value of these genes. In addition, mRNA expression levels of candidate diagnostic genes were analyzed in control samples and renal fibrosis samples. CIBERSORT algorithm was used to evaluate immune cells level. Additionally, gene set enrichment analysis (GSEA) and drug sensitivity were conducted.
After obtaining a total of 24 DEGs, we discovered that they were mostly involved in several immunological and inflammatory pathways, including NF-KappaB signaling, AGE-RAGE signaling, and TNF signaling. Five genes (COL4A2, CXCL1, TIMP1, VCAM1, and VEGFA) were subsequently identified as biomarkers for renal fibrosis through machine learning, and their expression levels were confirmed by validation cohort data sets and
COL4A2, CXCL1, TIMP1, VCAM1, and VEGFA are promising diagnostic biomarkers of tissue and serum for renal fibrosis.