Population Genomic Analysis of 962 Whole Genome Sequences of Humans Reveals Natural Selection in Non-Coding Regions

Abstract

Whole genome analysis in large samples from a single population is needed to provide adequate power to assess relative strengths of natural selection across different functional components of the genome. In this study, we analyzed next-generation sequencing data from 962 European Americans, and found that as expected approximately 60% of the top 1% of positive selection signals lie in intergenic regions, 33% in intronic regions, and slightly over 1% in coding regions. Several detailed functional annotation categories in intergenic regions showed statistically significant enrichment in positively selected loci when compared to the null distribution of the genomic span of ENCODE categories. There was a significant enrichment of purifying selection signals detected in enhancers, transcription factor binding sites, microRNAs and target sites, but not on lincRNA or piRNAs, suggesting different evolutionary constraints for these domains. Loci in “repressed or low activity regions” and loci near or overlapping the transcription start site were the most significantly over-represented annotations among the top 1% of signals for positive selection.

Sheep genome functional annotation reveals proximal regulatory elements contributed to the evolution of modern breeds

“…Second, levels of nucleotide diversity within each predicted annotation feature were compared. This showed epigenomic modifications reporting proximal promoters contained lower diversity than those marking distal enhancers, consistent with findings for other species 48 (Supplementary Fig. 10 ).…”

“…The reciprocal BLAST methodology may have introduced bias, resulting in an over-representation of evolutionarily conserved promoter elements compared with enhancers. Indeed the recovery of enhancers within sheep ChIP-Seq data was lower than for promoters, however the difference was not large and enhancers have a higher evolutionary turnover compared to proximal regulatory elements or protein-coding sequences 48 , 62 , 66 . Further, it is possible that species specific and tissue specific regulatory elements are likely to be under represented as a consequence of both our methodology and the application of ChIP-Seq using only a single tissue.…”

Genomic landscape of the signals of positive natural selection in populations of Northern Eurasia: A view from Northern Russia

“…Similar to the SNPs from the microarray part, all 57 SNPs with the strongest signals for selection from Tables 3 and 4, which were found by wholegenome sequence examination, are either in the middle of very large introns or inside intergenic regions. Such priority of location of SNPs under selection corresponds well with the data of other publications, which demonstrated that selection signals are pervasive in non-coding regions [51]. This suggests that selection preferably operates through modulatory and regulatory mechanisms, instead of changes in protein structure.…”

A hybrid computational strategy to address WGS variant analysis in >5000 samples

“…The yield of SNVs from our pipeline is less than that of 1000GP3 (~84 million) [ 1 ] even though the number of samples is almost twice as large, but that can be attributed to the relatively homogeneous ancestry of our samples compared with 1000GP3. In a previous paper [ 30 ] on ~1000 samples from our dataset, the yield of SNVs was ~ 24 million, which is less than that of a comparable sample size of 1000Genomes Phase 1 project. Even the UK10K SNV callset [ 2 ] has only ~42 million SNVs whereas the number of unrelated samples in that project exceeds that of 1000GP3.…”