Biobank Japan (BBJ)

We used two sets of GWAS summary statistics from BBJ for the EAS GWAS meta-analysis and the SNP-based heritability and genetic correction analyses, respectively. For the EAS GWAS meta-analysis, we used the latest BBJ GWAS summary statistics from Sakaue et al. (https://pheweb.jp/).^5,12,14 Genome-wide association tests were performed for 220 phenotypes, using linear or logistic mixed models implemented in SAIGE⁴⁶ (v0.37) or BOLT-LMM⁴⁷ (v.2.3.4) adjusted for age, age,² sex, age by sex interaction, age² by sex interaction, and the top 20 PCs. We extracted 24 quantitative traits from BBJ that matched the quantitative traits we analyzed in TWB: height (HT), weight (WT), body mass index (BMI), diastolic blood pressure (DBP), systolic blood pressure (SBP), white blood cell (WBC), red blood cell (RBC), hemoglobin (HB), hematocrit (HCT), platelet (PLT), blood urea nitrogen (BUN), creatinine (CR), uric acid (UA), total bilirubin (T-BIL), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), albumin (ALB), fasting glucose (FG), hemoglobin A1c (HbA1c), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG). We meta-analyzed these 24 GWAS in BBJ with the corresponding GWAS in TWB using sample size-weighted Z score meta-analysis implemented in an in-house script (Python 3.9.5).¹⁹ We retained variants presented in either TWB or BBJ in this meta-analysis for loci discovery in the East Asian populations. Phenotypes used in the BBJ GWAS were either inverse rank-based normal transformed (HT, WT, BMI, T-BIL, ALB, FG, HbA1c), log transformed followed by Z score transformed (BUN, CR, ALT, AST, GGT, TG), or Z score transformed (DBP, SBP, WBC, BRC, HB, HCT, PLT, UA, TC, HDL-C, LDL-C).

For SNP-based heritability and genetic correction analyses, we used an earlier release of BBJ GWAS summary statistics from Kanai et al. (http://jenger.riken.jp/en/result),¹³ because association test statistics from mixed models can bias heritability and genetic correlation estimates when used with LD score regression.⁴⁸ We extracted BBJ GWAS summary statistics for 20 traits from this earlier release, including BMI, DBP, SBP, WBC, RBC, HB, HCT, PLT, CR, T-BIL, ALT, AST, GGT, ALB, FG, HbA1c, TC, HDL-C, LDL-C, and TG. The association tests in these earlier BBJ GWAS were performed by first residualizing phenotypes on age, age,² sex, the top 10 PCs, and trait-specific covariates (e.g., disease status), followed by association tests in linear regression.