The temporal and genomic scale of selection following hybridization

Abstract

Genomic evidence supports an important role for selection in shaping patterns of introgression along the genome, but frameworks for understanding the evolutionary dynamics within hybrid populations that underlie these patterns have been lacking. Due to the clock-like effect of recombination in hybrids breaking up parental haplotypes, drift and selection produce predictable patterns of ancestry variation at varying spatial genomic scales through time. Here, we develop methods based on the Discrete Wavelet Transform to study the genomic scale of local ancestry variation and its association with recombination rates and show that these methods capture temporal dynamics of drift and genome-wide selection after hybridization. We apply these methods to published datasets from hybrid populations of swordtail fish (Xiphophorus) and baboons (Papio) and to inferred Neanderthal introgression in modern humans. Across systems, upward of 20% of variation in local ancestry at the broadest genomic scales can be attributed to systematic selection against introgressed alleles, consistent with strong selection acting on early-generation hybrids. Signatures of selection at fine genomic scales suggest selection over longer time scales; however, we suggest that our ability to confidently infer selection at fine scales is likely limited by inherent biases in current methods for estimating local ancestry from contiguous segments of genomic similarity. Wavelet approaches will become widely applicable as genomic data from systems with introgression become increasingly available and can help shed light on generalities of the genomic consequences of interspecific hybridization.

Keywords: Neanderthal; hybridization; introgression; wavelet transform.

Fig. 1.

(Left) Ancestry states $x(\ell )\in \{0,1\}$ along three hypothetical chromosomes, with examples of Haar wavelets overlaid in dark gray. From Top to Bottom, ancestry tracts are shorter, representing different histories of recombination. Shaded intervals highlight the portion of the ancestry signal contributing to the resulting wavelet coefficient. (Top Left) Positive covariance between a ${\psi }_{\lambda =6}$ wavelet and x within the shaded interval yields a positive wavelet coefficient corresponding to a change in average ancestry state over the two halves of the chromosome. (Middle Left) Negative covariance between a ${\psi }_{\lambda =4}$ wavelet and x within the shaded interval yields a negative wavelet coefficient. (Bottom Left) Positive covariance between a ${\psi }_{\lambda =2}$ wavelet and x within the shaded interval gives a positive wavelet coefficient. (Right) The complete set of squared wavelet coefficients determines the power spectrum for the three ancestry signals, correspondence indicated in color.

Fig. 2.

Wavelet variance decomposition through time of ancestry proportion in a 50/50 population mixture undergoing genetic drift with recombination and no selection. We simulated using SLiM, (41) a population of constant size 2N = 20,000 (blue) and a population that undergoes a bottleneck to 2N = 200 for just the first 10 generations of recombination in hybrids, then expands to 2N = 20,000 (maroon). (Top) Ancestry proportion along human chromosome 1 from a single simulation run. From left to right, shown after 10, 100, and 1,000 generations of recombination in hybrids. (Bottom) Wavelet variance decomposition showing the spatial scale of variance in ancestry proportion. Points and error bars show means and 95% CIs across 20 replicate simulations. Solid gray lines show theoretical expectations. Vertical dotted gray lines indicate the expected distance between recombination breakpoints that have accrued along a single chromosome since the hybridization pulse. Note that since results are shown on the genetic map, recombination rate variation does not influence these patterns, other than through interpolation error.

Fig. 3.

Simulations of selection following a pulse of hybridization starting from a 50/50 population mixture. (A) Genome-wide selection in conjunction with broad-scale variation in recombination rates leads to differential removal of introgressed ancestry at broad scales, thereby biasing the power spectrum toward greater variance at these scales compared to the neutral expectation (viewed after 100 generations). (B) Selection against many alleles on one ancestry background rapidly establishes broad-scale correlations between recombination and minor parent ancestry in early generation hybrids. (C) Through time with continued selection, the overall correlation remains dominated by broad scales, but finer scales contribute increasingly more. (D) Selection acting only on the first 10 generations of recombinant hybrids generates significant positive wavelet correlations only at broad scales (brown) (viewed in generation 1000), whereas continuous selection over 1000 generations continues to generate correlations on finer scales (red). When selection acts continuously but reverses direction after 100 generations to favor the alternate ancestry, positive broad-scale correlations persist as negative correlations establish at finer scales (see also SI Appendix, Fig. S7B). Only significant correlations are shown; error bars represent 95% CIs across 20 replicate simulations.

Fig. 4.

Wavelet analysis of hybrid genomes in a population of X. birchmanni×X. malinche swordtail fish from Acuapa River in Hidalgo, Mexico. (A) Power spectrum of the malinche-like ancestry proportion for five time points between 2006 and 2018. Points are weighted averages across chromosomes, and error bars are 95% jackknife CIs. (B) Correlations at each spatial genomic scale (on the physical map) between wavelet coefficients for the proportion of malinche-like haplotypes and wavelet coefficients for recombination rate. Squared values give an estimate of the proportion of variance in ancestry state explained by systematic selection against malinche-like alleles. Data shown only for 2006, patterns similar across years. (C) Contribution of each genomic scale to the overall correlation.

Fig. 5.

Wavelet analysis of hybrid genomes between yellow and anubis baboons in Amboseli, Kenya. (A) Power spectrum of the proportion of anubis-like haplotypes within diploids on the genetic map, stratified by quintile of genome-wide average anubis-like ancestry. Error bars are 95% CIs using the SE of the wavelet variance across individuals within each quintile. (B) Correlations at each spatial genomic scale (on the physical map) between wavelet coefficients for sample proportion of anubis-like haplotypes and wavelet coefficients for recombination rate. Squared values give an estimate of the proportion of variance in ancestry state explained by systematic selection against anubis-like alleles. Weighted-jackknife 95% CIs shown for all but the largest scale which is present only on a single chromosome. (C) Contribution of each scale to the overall correlation. Although there is a strong negative correlation at scale 102,400 kb, this scale is only present on chromosome 1 and does not contribute substantially to the overall positive correlation.

Fig. 6.

Wavelet analysis of inferred Neanderthal ancestry in modern humans. (A) Normalized power spectra of the proportion of Neanderthal-like tracts for three different studies (colored points) compared to a neutral expectation for 2,000 generations of admixture in a population size of 10,000 diploids (gray line). Error bars represent 95% CIs from a weighted jackknife across chromosomes. (B) Correlations across scales between different similarity calls from different datasets. (C) Contribution of each scale to the overall observed correlation between the proportion of Neanderthal-like haplotypes and log-transformed recombination rates. Shapes correspond to different studies as indicated in panel (A).