Abstract
Abstract Background Genetic risk for Alzheimer Disease (AD) varies across populations with different ancestry. Three‐dimensional (3D) genome architecture regulates gene transcription. Genome‐wide Hi‐C analyses have revealed dynamic 3D genome changes during development and disease progression. The variation of 3D genome architecture on different ancestral background, however, has not been extensively studied, especially at high‐resolution level. Recently, we developed the DeepLoop pipeline enabling robust high‐resolution analysis. We applied DeepLoop to characterize 3D genome architecture in frontal cortex from African American (AA) and European (EU) donors while considering local ancestry (LA) in admixed samples. Methods Hi‐C libraries were constructed and ∼500 million of paired‐end reads were obtained in 8 samples (4 AA and 4EU) that were sex‐matched and homozygous for APOEε4. We partitioned Hi‐C data into compartment A (open and transcriptionally active chromatin) and compartment B (compacted and transcriptionally silenced chromatin). In total, 6206 compartment bins across the genome were compared at a 500kb resolution. DeepLoop was used to call chromatin loop (often represents promoter‐enhancer interaction) at a 5kb‐resolution. LA was calculated using RFMix with SNP‐array data and a reference panel from 1000 Genomes for EU, African (AF) and Amerindian (AI) ancestry. Results Most of the A and B compartments were consistent among individuals with only <5% of them switched between AA and EU individuals. The EU samples have slightly (107 bins) more transcriptionally active bins than the AA samples. While many of homozygous LA blocks displayed same chromatin loop pattern between and within ancestries, we did observe ancestry‐specific loop as well as differences between individuals of the same ancestry. Importantly, the presence of heterozygous LA (African/EU) at some loci displayed a composite loop pattern that was present in the same region of homozygous LA of African or EU ancestry. Conclusion The 3D genome architecture can be variable between different ancestral backgrounds as well as between individuals of the same LA. We hypothesize that admixture of LA can create novel or demolish chromatin interactions observed in the homozygous LA. These data emphasize the importance of LA in identifying the causal loci contributing to AD in admixed population such as AF and Hispanic/Latino.