Multi‐Ancestry Genetics for Alzheimer’s Disease: Alzheimer’s Disease Genetics Consortium

Abstract

AbstractBackgroundAlzheimer’s disease (AD) affects individuals of all races and ethnicities, yet large‐scale genetic research to compare genetic makeups for risk or protection of AD between multi‐ancestry datasets is limited. The Alzheimer’s Disease Genetics Consortium (ADGC) assembled the largest multi‐ancestry dataset including 56,241 subjects from non‐Hispanic Whites (NHW), African Americans (AA), Hispanics (HI), and East Asians (EAS) to understand ancestral similarities and differences in genetic risk and protective variants for AD.MethodWe established a Steering Committee with 5 members, 4 main Workgroups including Primary Meta‐Analysis, Follow‐Up Analysis, Secondary Analysis, and Method Evaluation, and 3 affiliated Workgroups consisting of Phenotype Harmonization, Imputation Development, and genome‐wide association study (GWAS)/whole genome sequencing (WGS) Integration. A new computing infrastructure was developed for shared analytic space on Amazon Web Services (AWS), where all members from different institutes accessed to genetic and phenotypic data and conducted analysis under the master analysis plan with centrally developed pipelines and scripts. We conducted ancestry‐specific GWAS using single nucleotide polymorphisms (SNPs) for AD, cross‐ancestry meta‐analysis of 4 ancestry GWASs, fine‐mapping for allelic heterogeneity in known and novel loci for AD, and pathway analysis.ResultWe confirmed 12 known including APOE ɛ4 and ɛ2 variants and identified 2 novel loci for AD in cross‐ancestry meta‐analysis with P<5.4×10−8. African and Asian ancestry subjects displayed greater heterogeneity in effect sizes and significant levels with the previously known AD loci for AD. Ancestry‐specific fine‐mapping indicated presence of potential distinct causal variants across different ancestry groups from multiple known AD loci. Two APOE alleles had greatest effect sizes among East Asians compared to others (EAS>NHW>AA>HI), while we failed to observe significant association signals with AD in SHARPIN and TREM2 regions from African Americans. Of two novel loci, the association signal in the LRRC4C‐API5 region was contributed from all four ancestry groups; in contrast, the signal in the LHX5‐AS1‐RBM19 region was predominantly from Hispanics.ConclusionGenetic risk profiling in multi‐ancestry datasets provided unique opportunity to understand ancestry‐specific and cross‐ancestry genetic risk factors for AD. Further evaluation of these established and novel loci for AD by integrating GWAS and WGS data from ADSP is underway