Two-dimensional site frequency spectrum for detecting, classifying and dating incomplete selective sweeps

Abstract

The two-dimensional site frequency spectrum (2D SFS) was investigated to describe the intra-allelic variability (IAV) maintained within a derived allele (D) group that has undergone an incomplete selective sweep against an ancestral allele group. We observed that recombination certainly muddles the ancestral relationships of allelic lineages between the two allele groups; however, the 2D SFS reveals intriguing signatures of recombination as well as the genealogical structure of the D group, particularly the size of a mutation and the time to the most recent common ancestor (TMRCA). Coalescent simulations were performed to achieve powerful and robust 2D SFS-based statistics with special reference to accurate evaluation of IAV, significance of recombination effects, and distinction between hard and soft selective sweeps. These studies were extended to a case wherein an incomplete selective sweep is no longer in progress and ceased in the recent past. The 2D SFS-based method was applied to 100 intronic linkage disequilibrium regions randomly chosen from the East Asian population of modern humans to examine the P value distributions of the summary statistics under the null hypothesis of neutrality in a nonequilibrium demographic model. We argue that about 96% of intronic variants are non-adaptive with a 10% false discovery rate. Furthermore, this method was applied to six genomic regions in Eurasian populations that were claimed to have experienced recent selective sweeps. We found that two of these genomic regions did not have significant signals of selective sweeps, but the remaining four had undergone hard and soft sweeps and were dated, in terms of TMRCA, after the major out-of-Africa dispersal of modern humans.

Population genomics on the origin of lactase persistence in Europe and South Asia

“…The hallmark of an incomplete selective sweep is reduced intra-allelic variability (IAV) within a derived allele group relative to the total variability in the entire sample (Fujito et al, 2018;Satta et al, 2019). IAV may be expressed by numerous variables that are useful for detecting and classifying a selective sweep as well as for dating the onset of positive selection.…”

“…Indeed, almost all alleles per SNP are singletons or doubletons, irrespective of populations, such that HJ is close to 1 to 2. No other target sites that exhibit such low multiplicity in such a wide genomic region have yet to be revealed (Satta et al, 2019). The 2D SFS method also allows us to estimate the mutation-rate-scaled TMRCA ( "#$ ) from the total number ( = ∑ -,J /…”

“…To detect, classify and date an incomplete selective sweep for LCT in each population, we used the method of the two-dimensional site frequency spectrum { -,. }, 2D SFS, by Satta et al (2019). This method first seeks for a region under strong linkage disequilibrium (LD with 1 > 3/4) surrounding a target site and divides a sample of size into the derived and ancestral allele groups defined by the site.…”

“…We may compare the performance of our inference method with that of recent studies. Such studies include Itan et al (2009) who estimated 6256 < "#$ < 8683 using an approximate Bayesian computation (ABC) approach, Field et al (2016) who developed the singleton density score method and estimated "#$ ≥ 2000 from the UK10K project, Smith et al (2018) who developed a hidden Markov model and obtained "#$ = 9341 (95% CI: 8688-19,989) for IBS and 9514 (95% CI: 8596-10,383) for PJL, Harris et al (2018) who based their method on haplotype structure, Akbari et al (2018) who claimed the high power of their iSAFE method to pinpoint the causative mutation of a single sweep, Nakagome et al (2019) who developed an ABC method conditioned on the allele frequency in the past and applied it to four pigmentation SNPs, Satta et al (2019) who used the 2D SFS method for EUR and estimated "#$ = 3280 ± 480, and Tournebize et al (2019) who obtained "#$ = 4250 (95% CI: 3700-17,680) for CEU. Readers may also refer to Tournebize et al…”

“…We assumed the size of the bottlenecked population to be 2000 and ignored the effect of population explosion during the last 500 years. The robustness of the method to nonequilibrium demography was discussed in Satta et al (2019). For genome evolution, we assumed the infinite-sites model (Kimura, 1969), but with linkage.…”