Integrative transcriptomic and machine-learning analysis reveals shared molecular signatures linking osteoporosis and atherosclerosis

Osteoporosis (OP) and atherosclerosis (AS) frequently coexist with ageing, yet their shared molecular features remain incompletely understood. We separately analyzed femoral-bone OP and arterial AS transcriptomes using Benjamini–Hochberg-adjusted differential expression. Arterial-territory adjustment, weighted gene co-expression network analysis (WGCNA), and repeated-resampling machine learning were applied to the AS cohort, alongside functional enrichment and STRING network analyses. Intersecting OP differentially expressed genes with AS-associated WGCNA modules identified 2788 shared candidates, of which 88 also met stringent differential-expression criteria in both diseases. Functional and network analyses highlighted translation, endosomal–lysosomal biology, endocytosis, sphingolipid metabolism, myeloid-associated processes, and extracellular-matrix remodeling. C1orf174, FBXL18, and RBM4B were consistently prioritized by LASSO, random forest, and SVM-RFE. A fixed nine-gene candidate panel was evaluated in independent circulating-transcriptome cohorts. In GSE56815, C1orf174 met targeted panel-level correction (FDR = 0.0429), but not genome-wide correction. Exploratory clinical RT-qPCR showed nominally lower FBN1 expression in the OP + AS group, whereas PTEN and SMAD2 showed no statistically significant differences. These findings provide an integrative framework for the bone–vascular axis and prioritize candidate genes for further mechanistic investigation and independent validation.