Machine learning reveal shared diagnostic biomarkers and convergent pathways in age-related hearing loss and sarcopenia.

Age-related hearing loss (HL) and sarcopenia (ARS) are prevalent geriatric syndromes sharing common risk factors. This study aimed to identify shared biomarkers and elucidate convergent pathogenic mechanisms. Transcriptomic datasets were obtained from public database. Differential expression analysis was performed, followed by enrichment analysis. Hub genes were identified via LASSO regression, SVM-RFE, and random forest algorithms. Diagnostic performance was evaluated using receiver operating characteristic curve analysis across 6 independent cohorts. Comprehensive integrative analysis revealed distinct yet overlapping molecular signatures between HL and ARS. In HL, 11 upregulated and 16 downregulated genes were shared between 2 diseases, and complement and coagulation cascades, Toll-like receptor signaling, efferocytosis, as well as immune response processes were found to be associated with these genes. Machine learning identified 10 hub genes (AIMP2, JUN, SEMA5A, RASL12, GUSB, C1QA, GYPC, IRF7, C1QB, SERPING1) as shared biomarkers. Notably, these genes demonstrated robust diagnostic utility: individual genes exhibited area under the curve (AUC) values > 0.7 in most cohorts. Although the combined 10-gene model achieved AUC = 1 in several cohorts, these results should be interpreted with caution due to the limited sample sizes in some datasets (e.g., GSE6045, n = 3 per group), which may inflate performance metrics. Permutation tests confirmed that the AUC values were significantly better than chance in several cohorts (P < .05). This study pioneers a machine-learning framework to uncover shared molecular drivers of HL and ARS, identifying 10 hub genes as promising diagnostic biomarkers.