Abstract
Abstract Background Lack of diversity in expanding genomic studies has limited the identification of LOAD risk variants that may be more prevalent in non‐European ancestry groups with more modest power requirements for detection. To help reverse this trend, we constructed and analyzed a multi‐ancestry collection of GWAS datasets in the Alzheimer’s Disease Genetics Consortium (ADGC) to identify novel shared and ancestry‐unique LOAD susceptibility loci and related biological pathways. Method The multi‐ancestry data collection of the ADGC includes GWAS genotype and phenotype data on 38,774 non‐Hispanic White (NHW), 7,454 African American (AA), 11,436 Hispanic (HI), and 3,277 East Asian (EAS) subjects imputed to the NHLBI TOPMed v5 reference panel. We performed a two‐stage analysis: (1) single‐variant association using score‐based logistic regression for population‐based datasets and generalized linear mix‐models for family studies with covariate adjustment for onset/exam age, sex, principal components for population substructure, and APOE ε2/ε3/ε4 genotype, followed by within‐ancestry fixed‐effects meta‐analysis using METAL; and (2) cross‐ancestry meta‐analysis of within‐ancestry associations using random‐effects modeling (RE2) in METASOFT. Secondary analyses include pathway analysis using EnrichR and gene‐based association testing using SKAT‐O (on‐going). Result In addition to APOE region associations, we identified 13 loci with genome‐wide significant (GWS; P≤5×10 −8 ) cross‐ancestry associations including chromosomes 1q32.2 (CR1), 2q14 (BIN1), 6p21.1 (TREM2), 6p12.3 (CD2AP), 8p21.2 (PTK2B), 8p21.1 (CLU), 8q24.3 (SHARPIN), 11q12 (MS4A6A), 11q14 (PICALM), 19p13.3 (ABCA7), and novel loci at 11p2 (LRRC4C) and 12q24.13 (LHX5‐AS1). An additional 13 loci reached suggestive significance (P<10 −5 ), including loci with ancestry‐specific associations attaining GWS such as PALM2AKAP2 (P = 2.5×10 −8 in EAS), GRB14 (P = 1.7×10 −8 in HI), and KIAA0825 (P = 2.9×10 −8 in NHW). Follow‐up pathway analysis implicated multiple amyloid regulation pathways (strongest: negative regulation of APP catabolism, adjusted P = 1.6×10 −4 ) and the classical complement pathway (adjusted P = 1.3×10 −3 ). Follow‐up analyses including fine‐mapping and gene‐based analyses are on‐going. Conclusion Cross‐ancestry GWAS meta‐analyses identified novel LOAD susceptibility loci in/near LRRC4C and LHX5‐AS1, both with known roles in neuronal development, as well as several novel ancestry‐unique loci, showing their tremendous value for gene discovery with smaller sample sizes than current European ancestry LOAD GWAS. Even larger multi‐ancestry studies will provide even more power for further elucidating the genomics of LOAD.