Identification and initial experimental assessment of DCLK3 as a potential sarcopenia biomarker: A multi-omics and machine learning approach

Sarcopenia is an age-related condition marked by high prevalence and significant clinical consequences, underscoring the need for timely diagnostic and predictive strategies. In this study, we applied Weighted Gene Co-expression Network Analysis (WGCNA) to detect gene modules associated with sarcopenia, followed by predictive modeling using machine learning. This methodology aimed to identify robust biomarkers to support early detection and tailored therapeutic strategies. We integrated three sarcopenia-related gene expression datasets from the GEO database. Functional enrichment analysis of differentially expressed genes (DEGs) was performed to interpret their involvement in disease mechanisms. Key modules and hub genes were discerned through WGCNA and machine learning algorithms. Expression patterns of candidate genes were confirmed via RT-qPCR in sarcopenia samples. Gene Set Enrichment Analysis (GSEA) helped uncover signaling pathways linked to DCLK3 expression. Single-nucleus RNA sequencing (snRNA-Seq) was employed to assess transcriptional heterogeneity and cell-cell communication networks during disease progression. Our analysis revealed 472 DEGs and 1325 highly correlated genes from WGCNA. Among 24 overlapping genes, three machine learning approaches—LASSO, Random Forest, and SVM-RFE—converged on DCLK3 as a central biomarker. RT-qPCR confirmed marked dysregulation of DCLK3 in sarcopenia. snRNA-Seq and communication analysis further showed enriched DCLK3 expression in pericytes, with PDGF signaling as the major pathway mediating their crosstalk with other cells. These results position DCLK3 as a compelling diagnostic marker and therapeutic target, providing new insights into sarcopenia pathogenesis.