Identification of haplotype blocks using character-compatibility analysis and inference of deep human history

Abstract

BACKGROUND: Recombination disrupts the inheritance of human chromosomes. As a result, at a population-level chromosomes exhibit a block-like structure ('haploblocks'). Each haploblock is an independent realisation of the evolutionary forces that shaped that population. Accurately partitioning the genome into haploblocks is difficult; partitioning using linkage disequilibrium (LD) often results in haploblocks with recombinant haplotypes that confound demographic reconstructions. To mitigate recombinant noise, I partitioned human chromosomes using a phylogenetically robust approach and subsequently reconstructed human demographic history.

METHODS: I identified haploblocks using character-compatibility (CC) matrices and various LD matrices on a test population. Identified haploblocks were validated post-hoc. I compared the structure of haploblocks identified using CC to those identified using LD. The impact of recombination on evolutionary inferences was compared using population genetic summaries, inferred haplotrees, and inferred temporal changes in population size per haploblock. I compared the distribution of these evolutionary inferences for haploblocks with and without the presence of recombinant haplotypes and between methods. I then identified haploblocks in 26 geographically diverse populations using CC using the same approach. I also compared demographic histories inferred using my CC approach to commonly-used methods of demographic inference.

RESULTS: Using CC to identify haploblocks was more accurate than using LD. Using CC, I identified a greater number and proportion of the genome without recombination. Haploblocks without recombination were shorter than haploblocks with recombination; ~2kbp vs. ~20kbp. Population genetic summaries were mildly different when recombinant haplotypes were present. Recombination did not affect the estimated time to most recent comment ancestor (TMRCA) of haplotrees, which all have a median TMRCA of ~1.5ma. Pleistocene population bottlenecks are evident in haploblocks regardless of recombination; however, recombination lessens the severity of the bottleneck and the contemporary population sizes are larger. The haploblocks in different human populations have a median length of 1,000-3,000bp. African populations were partitioned into smaller haploblocks compared to non-African populations. Population genetic inferences were congruent with those reported elsewhere in the literature. The median TMRCA of haplotrees was ~1.5 million years ago. Non-African populations exhibit a distinct population bottleneck beginning 150,000 years ago. Non-African populations tend to exhibit a greater contemporary effective population size compared to African populations. The mid-late Pleistocene bottleneck I detected in non-Africans was present in most of the other commonly-used gold standard demographic reconstruction methods. African populations exhibited little concordance between any method.

SIGNIFICANCE: My results demonstrate that using CC to identify haploblocks without the presence of recombinant haplotypes in human genomes is more accurate than LD. This is pertinent when recombination will be a confounding factor, such as using haploblocks for evolutionary analyses. However, my results suggest that recombination has differing effects depending on the analyses being undertaken. My results show differences in global haploblock structure that is congruent with the literature. Non-African haploblocks are likely longer due to them having emerged from a severe population bottleneck in the mid-late Pleistocene. My results also suggest that on average human genealogical lineages pre-date the emergence of modern humans (300ka). These results expand upon previous studies, which were limited to ~40 loci. I have demonstrated that using haploblocks for evolutionary analyses can complement existing methods. Show more