In silico virtual knockout identifies PXDNL as a fibroblast-specific driver of sarcopenia and GABA as a potential modulator.

Sarcopenia lacks causal mechanisms and translatable targets. We integrated virtual gene knockout with multi-omics (n = 238 biopsies, 5 GEO cohorts) and single-cell RNA-seq (n = 10, 12,847 cells) to identify fibroblast-specific drivers. After ComBat batch correction, WGCNA identified a red module (690 genes, r = 0.74, P < 0.001) intersecting with 304 differentially expressed genes (|log2FC|>0.585, FDR<0.05), yielding 163 candidates. scTenifoldKnk virtual gene knockout ranked PXDNL as the top fibroblast-specific driver (perturbation score = 2.34, CV<15%), perturbing 327 ECM genes (e.g., FBN1 ΔE = +0.82, LRRTM4 ΔE = -0.71). A 12-gene panel (including PXDNL) was derived from 113 ML algorithm benchmark (plsRglm optimal: training AUROC = 0.938, external validation AUROC = 0.804, 95%CI:0.636-0.938). Drug repurposing (DSigDB, Z > 2.0) identified GABA as the top candidate. Molecular docking revealed strong PXDNL-GABA binding (ΔG = -5.6 kcal/mol) at the peroxidase domain, which was further validated by enzymatic activity assays. In dexamethasone-induced and TNF-α induced atrophy models of C2C12 or HMCs, 50 μM GABA restored cell viability (P < 0.001), downregulated Atrogin-1 (FBXO32)/MuRF-1 (TRIM63) (P < 0.01), and reversed PXDNL overexpression effects. This study establishes the fibroblast-PXDNL-ECM axis as a causal mechanism in sarcopenia and validates GABA as a repurposable therapeutic, providing a complete in silico-to-in vitro framework for age-related muscle disease.