Prioritizing Osteoporosis-Associated Candidates Through Multi-Omics Mendelian Randomization, Transcriptomic Modeling, and in vitro Validation

Background: Gut microbiota, immune traits, and circulating metabolites may contribute to skeletal biology. We integrated genetic association analyses, transcriptomic profiling, and cell experiments to prioritize osteoporosis-associated molecular candidates. Methods: Two-sample Mendelian randomization (MR), gene prioritization, and three GEO datasets comprising 110 samples were integrated. MR estimates were recalculated while retaining outcome-associated instruments, followed by false-discovery-rate correction. Exploratory classifiers and a five-gene nomogram were evaluated internally. Protein expression and osteoclast formation were examined in control and RANKL-treated RAW264.7 cells in three independent experiments per group. Supplementary validation repeated normalization, screening and tuning within training folds. Results: Among 2,329 IVW tests, the metabolite exposure indexed by GCST90200621 was associated with osteoporosis after multiplicity adjustment (odds ratio 1.296, 95% confidence interval 1.155– 1.454; within-domain adjusted P = 0.0135). No colocalization record had PP.H4 above 0.5. The transcriptomic workflow prioritized CCR1, TRMT2A, FADS2, FBXL14, and TTLL12. The original nine-gene RF analysis had an internal test-partition AUC of 0.805. Expression of all five proteins increased during RANKL-induced osteoclastogenesis. Additional nested validation yielded a mean AUC of 0.528; cross-study performance remained limited. Fixed-five-gene ridge sensitivity analysis yielded a mean internal AUC of 0.756, conditional on prior gene selection. Conclusion: Integrating genetic and transcriptomic analyses with cell experiments prioritized five osteoporosis-associated candidates whose protein expression increased during osteoclast differentiation. These complementary findings provide a focused basis for biological follow-up; predictive and gene-specific functional applications require further validation. Keywords: osteoporosis, mendelian randomization, gut microbiota, transcriptomics, machine learning, osteoclastogenesis