Identification of Energy Metabolism-Related Subtypes and Diagnostic Biomarkers for Osteoarthritis by Integrating Bioinformatics and Machine Learning.

Background: Osteoarthritis (OA) is a chronic and complex degenerative joint disease that increasingly burdens and affects the elderly population. Abnormal energy metabolism is closely associated with the pathological mechanisms of OA. This study aims to identify key genes related to energy metabolism that are closely linked to the treatment and diagnosis of OA. Methods: The transcriptomic data for OA were collected from the Gene Expression Omnibus (GEO), with GSE51588 and GSE63359 serving as the training and validation datasets, respectively. Differential expression analysis was conducted to identify key energy metabolism-related genes. Unsupervised clustering was performed to classify molecular subtypes. Three machine learning algorithms were employed to identify key diagnosis genes, specifically Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine Recursive Feature Elimination (SVM-RFE), and Random Forest (RF). We construct a comprehensive nomogram, and the diagnostic performance of both the nomogram and feature genes was evaluated using operating characteristic curve (ROC) and calibration curves. We assessed the immune infiltration levels in OA samples using the IOBR platform and the CIBERSORT algorithm. Results: We classified OA patients into two molecular subtypes, which exhibited distinct differences in immune infiltration levels. Subsequently, we successfully identified two characteristic genes, NUP98 and RPIA, and demonstrated statistically significant differences (P < 0.05) and diagnostic performance in the validation cohort. Evaluation using ROC and calibration curves demonstrated that these characteristic genes exhibit robust diagnostic performance. Multiple immune cells may be involved in the development of OA, and all characteristic genes may be associated with immune cells to varying degrees. Conclusion: In conclusion, NUP98 and RPIA have the potential to serve as diagnostic biomarkers for OA and may offer opportunities for therapeutic intervention in OA.