Long-read sequencing reveals hidden structural complexity within Alzheimers disease-associated GWAS loci

Abstract

Genome-wide association studies (GWAS) typically represent Alzheimers disease (AD) risk loci by a single lead SNP, although the lead SNP may capture only part of the genetic variation at a locus. We asked what genetic architectures underlie AD GWAS associations and how much information is obscured when a locus is reduced to a single lead SNP. We integrated AD GWAS summary statistics with long-read whole-genome sequencing of 493 individuals, phased DNA methylation, chromatin annotations, and structural variant (SV) imputation. Across 98 established AD loci, linkage disequilibrium (LD)-based clumping and conditional analysis identified 280 independent AD-haplotypes, of which 207 (74%) harbored linked structural variants, yielding 2,000 haplotype-linked SVs comprising intronic transposable elements and multiallelic tandem repeats within regulatory regions. Integration of methylation, chromatin annotations, and genomic context prioritized 52 AD haplotypes with candidate regulatory SVs, revealing that lead SNPs often capture only part of the underlying structural and regulatory architecture. At the SHARPIN/PLEC locus, long-read fine-mapping resolved an independent regulatory PLEC haplotype carrying a CpG-rich tandem repeat associated with enhancer methylation and reduced microglial expression. Extending these findings to population scale, we imputed 83% of haplotype-linked SVs into SNP-genotyped cohorts (median imputation R2 = 0.76). In exploratory model comparisons, lead-SNP-only models were favored at just 10% of loci, whereas models incorporating multiple haplotypes, SVs, or both were favored at the remaining 90%. Our findings show that AD GWAS loci frequently comprise structurally and regulatorily heterogeneous haplotypes that cannot be fully interpreted through the lead SNP alone. Rather than the lead SNP, the haplotype represents the fundamental unit for resolving the biological mechanisms driving AD genetic risk.