Abstract
BACKGROUND: Osteoarthritis (OA) is a prevalent chronic joint condition. This study sought to explore potential diagnostic biomarkers for OA and assess their relevance in clinical samples. METHODS: We searched the GEO database for peripheral blood leukocytes expression profiles of OA patients as a training set to conduct differentially expressed gene (DEG) analysis. Two machine learning algorithms, least absolute shrinkage and selection operator (LASSO) logistic regression and support vector machine-recursive feature elimination (SVM-RFE), were employed to identify candidate biomarkers for OA diagnosis. The performance was assessed using receiver operating characteristic (ROC) curves, and the areas under the curve (AUCs) with 95% confidence interval (CI) were calculated. Furthermore, we gathered clinical peripheral blood samples from healthy donors and OA patients (validation set) to validate our findings. Small interfering RNA and CCK8 proliferation assay were used for experimental verification. RESULTS: A total of 31 DEGs were discovered, and the machine learning screening found five DEGs that were considered to be candidate biomarkers. Notably, BIRC2 had a very good discriminatory effect among the five candidate biomarkers, with an AUC of 0.814 (95% CI: 0.697-0.915). In our validation set, results showed that the levels of BIRC2 and SEH1L were remarkably higher in healthy donors than OA patients, consistent with the results of the training set. SEH1L owned the largest AUC of 0.964 (95% CI: 0.855-1.000). BIRC2 also displayed a larger AUC of 0.836 (95% CI: 0.618-1.000) in the training set. Knockdown of these two genes could significantly suppress human chondrocyte proliferation. CONCLUSION: Two novel biomarkers, SEH1L and BIRC2, were indicated to have the capacity to differentiate healthy people from OA patients at the peripheral level. Experiments have shown that knockdown of these two genes could inhibit human chondrocyte proliferation, as verified by cell proliferation assays.