The study analyzed transcriptomic data from the GEO database and identified 808 differentially expressed genes in osteoarthritis (OA) synovium, of which 50 were related to ketone body metabolism. Using enrichment analysis, protein-protein network and machine learning, ACADL (long-chain acyl-CoA dehydrogenase) and ADH1B (alcohol dehydrogenase 1B) genes were selected as OA biomarkers. The nomogram based on these genes showed high accuracy in the training and validation sets. Functional analysis demonstrated that the genes are related to lipid oxidation, energy metabolism of synovial cells and redox balance against oxidative stress. Their expression was correlated with 21 immune cells, including pro-inflammatory M1 macrophages and Th17 cells. Molecular docking analysis identified progesterone and fomepizole as potential drugs binding to these genes. Mouse models of OA confirmed the reduction of ACADL and ADH1B proteins in the synovium. The findings link ketone body metabolic changes to immune inflammation in OA and support their use in diagnosis and targeted therapy.