Machine Learning Reveals the Diversity of Human 3D Chromatin Contact Patterns

• DOI: 10.1093/molbev/msae209
• Publication Date: 2024-10-15T10:57:48Z
• Authors: Erin N Gilbertson; Colin M Brand; Evonne McArthur; David C Rinker; Shuzhen Kuang; Katherine S Pollard; John A Capra
• Keywords: Divergence (linguistics); Sequence (biology)

Abstract Understanding variation in chromatin contact patterns across diverse humans is critical for interpreting noncoding variants and their effects on gene expression phenotypes. However, experimental determination of large samples prohibitively expensive. To overcome this challenge, we develop validate a machine learning method to quantify the 3D contacts at 2 kilobase resolution from genome sequence alone. We apply approach thousands human genomes 1000 Genomes Project inferred hominin ancestral genome. While divergence wide are qualitatively similar divergence, find substantial differences local 1 megabase genomic windows. In particular, identify 392 windows with significantly greater than expected sequence. Moreover, 31% windows, single individual has rare divergent map pattern. Using silico mutagenesis, that most nucleotide changes do not result contacts. just one or few can lead without individuals carrying those having high divergence. summary, inferring maps populations reveals variable patterns. anticipate these genetically will provide reference future work function evolution populations.