We examined MR associations between 2,923 UKB plasma proteins and osteoporosis using the inverse variance weighted (IVW) method. After false discovery rate (FDR) correction (q < 0.01), a total of 83 proteins showed significant causal associations with osteoporosis risk (Fig. 3 and Supplementary Table S1). Among them, 44 proteins were associated with increased osteoporosis risk (OR > 1), while 39 proteins were found to be protective (OR < 1). Notably, several proteins showed strong and consistent causal effects. For instance, GALNT3 (OR = 1.14, 95% CI: 1.10–1.19, p = 2.17 × 10⁻¹⁰), UBE2L6 (OR = 0.80, 95% CI: 0.75–0.87, p = 1.32 × 10⁻⁸), IL18 (OR = 1.17, 95% CI: 1.11–1.23, p = 1.46 × 10⁻⁸), NUDT2 (OR = 1.22, 95% CI: 1.13–1.31, p = 1.37 × 10⁻⁷), SMOC2 (OR = 0.90, 95% CI: 0.87–0.94, p = 2.57 × 10⁻⁷), TNFSF8 (OR = 1.27, 95% CI: 1.16–1.39, p = 3.91 × 10⁻⁷), RSPO3 (OR = 0.64, 95% CI: 0.53–0.76, p = 5.97 × 10⁻⁷), HEXIM1(OR = 0.45, 95% CI: 0.32–0.62, p = 2.07 × 10⁻⁶), CPA2 (OR = 0.90, 95% CI: 0.87–0.94, p = 2.64 × 10⁻⁶), IL7R (OR = 1.06, 95% CI: 1.04–1.09, p = 3.75 × 10⁻⁶), These findings suggest that certain plasma proteins may act as potential risk or protective factors in the pathogenesis of osteoporosis. The directionality and effect sizes were consistent across MR methods with no evidence of horizontal pleiotropy or heterogeneity in sensitivity analyses (Supplementary Table S2 and S3).

Causal associations between UKB plasma proteins and osteoporosis identified by two-sample MR

We performed MR analysis on deCODE plasma proteins to evaluate their causal roles in osteoporosis. After adjusting for multiple testing (FDR < 0.01), seven proteins were found to be significantly associated with osteoporosis risk (Fig. 4 and Supplementary Table S4). These included both risk and protective factors, with notable proteins as follows: GREM1 (OR = 1.21, 95% CI: 1.14–1.29, p = 1.37 × 10⁻¹⁰), PDE5A (OR = 1.28, 95% CI: 1.15–1.41, p = 2.08 × 10⁻⁶), PRRG4 (OR = 1.74, 95% CI: 1.36–2.23, p = 1.07 × 10⁻⁵), TMEM52B (OR = 1.29, 95% CI: 1.13–1.48, p = 1.51 × 10⁻⁴), BOLA1 (OR = 1.14, 95% CI: 1.07–1.22, p = 1.31 × 10⁻⁴), CCL19 (OR = 0.84, 95% CI: 0.77–0.92, p = 2.61 × 10⁻⁴), NT5C (OR = 0.78, 95% CI: 0.69–0.89, p = 1.90 × 10⁻⁴). Among these, GREM1 and PRRG4 displayed the strongest associations with increased osteoporosis risk, while NT5C and CCL19 showed protective effects. Further sensitivity analyses supported the reliability of these associations (Supplementary Table S5 and S6).

Causal effects of deCODE plasma proteins on osteoporosis

To investigate whether UKB plasma proteins causally influence the levels of deCODE plasma proteins, we conducted MR analysis using protein-matched instrumental variable datasets. Several significant protein-to-protein causal associations were identified (Fig. 5, Supplementary Table S7). Notable examples include: IL7R → NT5C (OR = 0.97, 95% CI: 0.96–0.99, p = 9.01 × 10⁻⁴), indicating a negative regulatory effect, GALNT3 → NT5C (OR = 1.03, 95% CI: 1.01–1.05, p = 0.0116), suggesting positive modulation. ATRN → NT5C (OR = 1.02, 95% CI: 1.00–1.04, p = 0.0242), SFTPA1 → NT5C (OR = 0.95, 95% CI: 0.91–0.99, p = 0.0321), CRELD1 → BOLA1 (OR = 1.01, 95% CI: 1.00–1.02, p = 0.0095), IL1RN → BOLA1 (OR = 0.95, 95% CI: 0.92–0.99, p = 0.0185), AHSG → GREM1 (OR = 1.06, 95% CI: 1.02–1.09, p = 7.87 × 10⁻⁴), PXN → GREM1 (OR = 1.02, 95% CI: 1.01–1.04, p = 0.0106), MMP12 → GREM1 (OR = 1.02, 95% CI: 1.00–1.04, p = 0.0266), IL18 → CCL19 (OR = 1.04, 95% CI: 1.01–1.07, p = 0.0049), NUDT2 → CCL19 (OR = 1.10, 95% CI: 1.05–1.15, p = 1.72 × 10⁻⁵), SMOC2 → CCL19 (OR = 0.97, 95% CI: 0.95–0.99, p = 0.0133), TGFBI → CCL19 (OR = 1.03, 95% CI: 1.01–1.05, p = 0.0111), IFNLR1 → TMEM52B (OR = 1.03, 95% CI: 1.00–1.06, p = 0.0225), B4GAT1 → TMEM52B (OR = 0.96, 95% CI: 0.93–1.00, p = 0.0398), These results suggest that several UKB proteins exert upstream regulatory effects on circulating deCODE proteins, many of which were also independently associated with osteoporosis. This protein-to-protein network supports the mechanistic foundation for further mediation analyses. Sensitivity analyses were also conducted to confirm this result (Supplementary Table S8 and S9).

Protein-to-protein causal relationships between UKB and deCODE plasma proteins

Two-step MR analyses were conducted to evaluate whether deCODE plasma proteins mediate the causal effects of UKB plasma proteins on the risk of osteoporosis. The results identified several significant mediating pathways. Specifically, NT5C was found to mediate the causal effect of GALNT3 and IL7R on osteoporosis, with mediation proportions of −5.51% and 9.80%, respectively. Likewise, BOLA1 mediated the relationships from CRELD1 and IL1RN to osteoporosis, accounting for 4.19% and 3.93% of the total effects, respectively. Notably, GREM1 emerged as a prominent mediator in multiple pathways. It mediated the causal associations from AHSG, MMP12, and PXN to osteoporosis, with mediation ratios of 17.95%, −6.98%, and 6.67%, respectively. In addition, CCL19 played a mediating role in the pathways linking IL18, NUDT2, SMOC2, and TGFBI to osteoporosis, with mediation ratios ranging from − 4.69% to −8.77%. These findings suggest that deCODE plasma proteins such as NT5C, BOLA1, GREM1, and CCL19 may partially account for the observed genetic effects of UKB plasma proteins on osteoporosis risk, providing mechanistic insights into disease pathogenesis. The mediation proportions are detailed in Figs. 6 and 7 and Supplementary Table S10.

Mediation effects of deCODE plasma proteins on the causal pathways between UKB plasma proteins and osteoporosis

Summary of significant mediation pathways between UKB plasma proteins and osteoporosis risk via deCODE plasma proteins