Machine learning-based identification of ferroptosis-related biomarkers in osteoarthritis

Abstract

Abstract Background Osteoarthritis (OA) is the most common joint disease and a major cause of chronic disability in elderly individuals. OA is characterized by degeneration of articular cartilage, structural changes in the subchondral bone structure, and formation of bony encumbrances, with the main clinical manifestations being joint swelling, pain, stiffness, deformity, and limited mobility. Ferroptosis is a newly identified form of lipid peroxidation-induced cell death. In recent years, several studies have shown that the pathological process of OA is related to ferroptosis. Objective The focus of this work was to identify and validate ferroptosis-related genes (FRGs) differentially expressed in osteoarthritis patients and to investigate potential molecular mechanisms. Methods The GSE98918 data were downloaded from the GEO database as the training set, and the GSE51588 data were used as the validation set. The differential gene expression of the training set was analyzed using R software and the ferroptosis-related differentially expressed genes. Then, machine learning algorithms were applied to build LASSO regression models and support vector machine models. After that, their intersection genes were used as feature genes to draw receiver operator characteristic (ROC) curves, and the resulting feature genes were validated using the validation set. In addition, the expression profiles of osteoarthritis were analyzed by immune cell infiltration, and the co-expression correlation between the characterized genes and immune cells was constructed. CONCLUSION KLF2 and DAZAP1 may serve as potential diagnostic biomarkers for OA. Meanwhile, KLF2 and DAZAP1 may be ferroptosis-related in OA, which provides insights for the development of new therapeutic approaches for OA.