Extended Data Fig. 6: Robustness of PGS to protein associations.

a-c) Robustness and longitudinal stability of PGS to protein associations to proteomics technology. d-e) Robustness and longitudinal stability of protein levels to proteomics technology. f) Robustness of PGS-protein associations to environmental and physiological confounding. g) Mediation of PGS-protein associations through body mass index (BMI) for six proteins associated with T2D PGS. a) Compares PGS-protein associations from Fig. 1b in n = 3,087 INTERVAL participants in which protein levels were measured with the SomaLogic platform (x-axis) to PGS-protein associations tested in an independent set of n = 418 INTERVAL participants in which protein levels were measured with the Olink Explore platform (y-axis). In total 1,463 proteins were quantified by the Olink Explore platform, including 907 quantified by the SomaLogic platform, and among these 16 of the 49 PGS-associated proteins from Fig. 1b. b) Compares PGS-protein associations from Fig. 1b (x-axis) to PGS-protein associations tested in an independent set of n = 3,848 INTERVAL participants in which protein levels were measured with the Olink T96 platform (y-axis). In total 265 proteins were quantified by the Olink T96 platform, including 224 quantified by the SomaLogic platform, and among these 4 of the 49 PGS-associated proteins from Fig. 1b. c) Compares PGS-protein associations tested in n = 646 INTERVAL participants in which protein levels were measured with both the SomaLogic platform (x-axis) and, after two-years of follow-up, the Olink T96 platform (y-axis). a-c) Data shown correspond to the beta estimates from linear regression (points) and their 95% confidence interval (bars), indicating standard deviation change in protein levels per standard deviation increase in the respective PGS (denoted by colour). Solid points indicate two-sided P-value < 0.05 for the test on the y-axis. Linear regression on both axes were adjusted for age (at protein measurement), sex, 10 genotype PCs, and platform-specific technical covariates. Full summary statistics including exact P-values are detailed in Supplementary Data 3,b for linear regression tests on y-axes, and in Supplementary Data 3,a for linear regression tests on x-axes. d) Compares protein levels quantified by the SomaLogic platform (x-axes) to protein levels quantified by the Olink T96 platform (y-axes) after two years of follow-up in n = 646 INTERVAL participants. e) Compares protein levels quantified by the Olink T96 platform (x-axes) to protein levels quantified by the Olink Explore platform (y-axes) in n = 418 INTERVAL participants. f) Compares PGS-protein associations from Fig. 1b in n = 3,087 INTERVAL participants (x-axes) to PGS-protein associations (1) additionally adjusted for circadian effects (time of day of blood draw), (2) additionally adjusted for seasonal effects (date of blood draw), (3) when including 87 additional participants with prevalent cardiometabolic disease (n = 3,174 on y-axis), and (4) when adjusting for BMI (n = 3,072 participants with non-missing BMI on y-axis). All associations were testing using linear regression adjusting for age, sex, 10 genotype PCs, sample measurement batch, and time between blood draw and sample measurement in addition to the covariates noted above. Data shown correspond to the beta estimates from linear regression (points) and their 95% confidence interval (bars), indicating standard deviation change in protein levels per standard deviation increase in the respective PGS (denoted by colour). Full summary statistics including exact P-values in these sensitivity analyses are detailed in Supplementary Data 3,c. g) For the six proteins whose association with T2D PGS was attenuated by adjustment for BMI (P > 0.05; Extended Data Fig. 6f) gives, from mediation analysis, the estimated effect of T2D PGS on the protein levels through BMI (standard deviation change in protein levels through BMI per standard deviation increase in T2D PGS), percentage of T2D PGS to protein levels mediated by BMI, and the estimated effect of T2D PGS on protein levels independent of BMI in n = 3,072 INTERVAL participants. All P-values are two-sided.