Abstract
BACKGROUND: Skeletal muscle aging is the major cause and hallmark of frailty, which poses a significant challenge to the healthcare system. AIM: This study aimed to identify the potential biomarkers for the early detection and therapeutic intervention of this age-related condition. METHODS: A transcriptomics-based methodology using machine learning algorithms was performed to select the biomarker genes. A predictive machine learning model for (pre-)frailty based on the transcriptomic profile of the biomarker genes was constructed and validated. The cell-type specific changes of the biomarkers during muscle aging were investigated in a single-cell RNA sequencing dataset of human skeletal muscle. Summary data-based Mendelian randomization (SMR) and Bayesian colocalization analyses were performed to identify biomarker genes with therapeutic effects on frailty-related skeletal muscle aging, and drug candidates were explored in the DSigDB database. RESULTS: We identified 24 biomarker genes, most of which were discovered for the first time. The optimal predictive model showed excellent performance in the external test set. Differential expression of the biomarkers in the single-cell dataset indicated a critical role of endothelial cells modulated by the marker genes MGP and ID1 in muscle degeneration. The SMR and colocalization analyses showed causal relationships between 2 marker genes (MGP and WAC) and frailty-related muscle aging. Potential therapeutics for MGP modulation were identified in the DSigDB database. CONCLUSIONS: This multi-omics study identified biomarkers associated with frailty-related muscle aging and provided new insights into the etiology and therapeutic targets for this age-related condition.