Screening of diagnostic markers related to immune infiltration in osteoarthritis patients based on machine learning

Abstract

Abstract Purpose We analyze the immune infiltration model of osteoarthritis to determine the relevant diagnostic biomarkers (OA), and to provide some help for the treatment and diagnosis of OA. Methods From the Gene Expression Omnibus (GEO) database, we downloaded GSE168505 and GSE114007 gene expression datasets, including 24 patients and 21 healthy controls. The R software Limma package and SVA package were used to analyze the batch effect. We selected differentially expressed genes (DEGs), and we then analyzed the DEGs’ functional enrichment. We performed differential analysis to pick out the differentially expressed immune-related genes (DEIRGs) in the merged data set. We first selected the candidate genes by the least absolute shrinkage and selection operator (LASSO) method, and then further screened the diagnostic markers by support vector machine-recursive feature elimination algorithm (SVM-RFE). In dataset GSE129147, the diagnostic value was determined by drawing the receiver operating characteristic (ROC) curve. In addition, we used the CIBERSORT program to assess the 22 kinds immune cells of infiltration models. Finally, an in vitro cell model of OA was established by interleukin-1β(IL-1β) to verify the bioinformatics results. Results Through differential analysis, 454 differential genes were identified, mainly involved ossification, extracellular matrix organization, collagen − containing extracellular matrix, metalloendopeptidase activity, PI3K − Akt signaling pathway, regulation of cell population proliferation, and other biological processes. We screened BIRC5 and TNFSF11 as candidate biomarkers by machine learning. In the data set GSE129147, BIRC5 and TNFSF11 were verified as diagnostic markers of OA by the ROC curve. The following correlation analysis found that BIRC5 and TNFSF11 were correlated with Mast cells resting, NK cells resting, Monocytes, Plasma cells, Eosinophil, Macrophages M0, and Macrophages M2. The expression of BIRC5 and TNFSF11 was up-regulated in the OA model in vitro. Conclusion We conclude that BIRC5 and TNFSF11 can be biomarkers for diagnosing OA. This discovery provides a direction for the occurrence of OA and the exploration of new treatment methods from the perspective of immunology.