Extended Data Fig. 8: The effect of population structure on metabolite clustering across datasets.

Principal component (PC) analysis was performed using joint metabolite data from the cohorts (AlphaOmega, n = 877; ERF, n = 778; RS1, RS Dataset 1, n = 2,975; RS2, RS Dataset 2, n = 729; RS3, RS Dataset 3, n = 1,487; TMS, n = 854). Two-tailed tests were used.