Fig. 4: GRS analyses in UKB and AoU.

a, In the UKB serology target cohort (n = 6,063, unrelated, EUR), EBVread⁺ status was predicted using a baseline model with or without one of six GRSs: imputed HLA alleles (HLA all); HLA alleles of MHC-I (HLA MHC-I); HLA alleles of MHC-II (HLA MHC-II); genotyped SNPs (SNP all); SNPs within MHC (SNP MHC); or all non-MHC SNPs (SNP wo MHC). Nagelkerke R² values are plotted (error bars denote standard deviations; bootstrapped, n = 1,000). b, GRSs were compared between sero⁻ (n = 348) and sero⁺ (n = 5,715) individuals and were non-significant (P values Bonferroni adjusted for four tests; statistical test: likelihood ratio test applied to logistic regression models, adjusted for covariates; see Methods). c, HLA all was positively correlated with EBV read counts. Sample sizes are indicated above the boxplots. d, Improvements in Nagelkerke R² for different AoU ancestry groups (HLA all GRS; abbreviations as in Extended Data Fig. 3), compared with the baseline model within UKB (from panel a). Error bars represent standard deviations (bootstrapped, n = 1,000). See the x axis for sample sizes. AFR, African; AMR, admixed American; EAS, East Asian; MID, Middle Eastern; SAS, South Asian. e, GRS distributions in individuals of the UKB serology target cohort and with EBV-associated diseases (UKB disease target cohort; see the x axis for sample sizes). Individuals with multiple diseases were included in each respective group. The statistical test is as in panel b. P values (Bonferroni adjusted) are provided if P < 0.1. ^aFor multiple sclerosis (MS), the signal was driven by HLA-A*02:01. HL, Hodgkin lymphoma; IM, infectious mononucleosis; NHL, non-Hodgkin lymphoma; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus. f, −log₁₀(P) of 1,751 PheCodes, grouped by organ systems or disease groups, for GRS HLA MHC-I and HLA MHC-II (statistical test as in panel b; dashed line: Bonferroni-corrected significance threshold) from the AoU QC EUR subset (n = 189,658). Phenotype terms are provided for test-wide significant results and for associations identified in panel e (encircled). IBD, inflammatory bowel disease; NOS, not otherwise specified; T1D, type 1 diabetes. The boxplots show the median (thick line), 25th and 75th percentiles (box) and the largest–smallest values no further from the box than 1.5 times the interquartile range (whiskers; b,c,e). Dashed lines in b,c,e correspond to values of 0.