Functional Annotation of Alzheimer's Disease-Associated SNPs Based on Three-Dimensional Chromatin Structure.

Genome-wide association studies (GWASs) have identified numerous loci associated with Alzheimer's disease (AD), yet the effector genes and regulatory mechanisms underlying many of these associations remain unresolved. This challenge is particularly pronounced for non-coding variants, whose regulatory effects may extend over long genomic distances and cannot be reliably inferred from the nearest gene alone. Here, we developed an integrative computational framework that incorporates three-dimensional chromatin interactions, expression quantitative trait loci (eQTL), and sequence-level regulatory annotations to identify candidate effector genes at AD-associated loci. AD-associated lead single-nucleotide polymorphisms (SNPs) and variants in strong linkage disequilibrium were mapped to enhancer elements and promoter-interacting regions using publicly available Hi-C and promoter capture Hi-C data from the hippocampus and cerebral cortex. Hi-C-derived enhancer-promoter relationships were inferred using PSYCHIC, whereas significant promoter-centered interactions from promoter capture Hi-C were used to connect variant-containing regions with candidate effector genes. The resulting SNP-gene relationships were further evaluated using brain-relevant eQTL evidence and predicted allele-dependent alterations in transcription factor binding motifs. This integrative analysis identified 608 unique candidate regulatory target genes supported by spatial chromatin contacts and complementary regulatory evidence, including genes not necessarily assigned by conventional nearest-gene annotation. Functional enrichment analysis indicated that the prioritized genes were involved in biological processes relevant to AD pathophysiology. Detailed analyses of three representative SNP-gene pairs, rs2373115-NARS2, rs6656401-CR1, and rs3776011-ACSL6, further illustrated how chromatin interaction, expression-associated, and sequence-level evidence can be combined to formulate locus-specific regulatory hypotheses. These findings represent computationally inferred and associative regulatory relationships rather than experimentally validated causal effects. They provide a structured framework for refining post-GWAS interpretation of non-coding AD risk loci and prioritizing candidate targets for further experimental validation.