Shared genetic architecture of osteoarthritis and bone traits (OA × eBMD × fracture): a five-layer cross-trait multi-omics funnel

Osteoarthritis (OA) and osteoporosis/fracture co-occur clinically, but whether their overlap reflects a unidirectional "low bone mass → OA" pathway or a shared bidirectional genetic architecture remains unresolved. We applied a five-layer cross-trait funnel (H1–H5) to OA, estimated bone mineral density (eBMD) and fracture. H1 (S-LDSC) revealed genome-wide correlation (rg(OA,eBMD) = +0.102; rg(OA,FX) = +0.171; rg(eBMD,FX) = −0.505). H2 (LAVA) localised 841/1,547 analysable regions at q < 0.05 (median ρ = 0.230), with 166 regions significant in all three pairwise tests against a permutation expectation of 143.9 ± 8.3 (P = 5.6×10⁻³). H3 prioritised 719 genes from 27 shared causal loci. H4 gene-level colocalisation found no gene reaching PP4 > 0.8 across 27 tissue-matched eQTL tests (max 0.200), indicating that the overlap is not mediated by a shared causal variant. H5 bidirectional Mendelian randomisation supported both bone→OA and OA→bone directions. This deposit contains all scripts, figures and result tables needed to reproduce the analysis.