Robust demographic inference from genomic and SNP data.

Excoffier L, Dupanloup I, Huerta-Sánchez E, Sousa VC, Foll M. · 2013

Read Robust demographic inference from genomic and SNP data. by Excoffier L, Dupanloup I, Huerta-Sánchez E, Sousa VC, Foll M. (2013) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3812088.txt.

Introduction

Laurent Excoffier, Isabelle Dupanloup, Emilia Huerta-Sánchez, Vitor C. Sousa, Matthieu Foll, and Joshua M. Akey present a new method paper in PLoS Genetics in 2013. Their tool infers past population history from gene data. It works from the site frequency spectrum, a summary of how gene variants are shared.

Their frame lets users test complex past events, such as population splits, size changes, migration, and hidden groups that left no samples. Tests show it stays fast and exact on large gene data sets. They apply it to SNP data from four human groups and to chip data on human origins. The method opens past events that older likelihood tools could not handle at all.

Catalog record

CREATORExcoffier L, Dupanloup I, Huerta-Sánchez E, Sousa VC, Foll M.
YEAR2013
COLLECTIONPAPERS
FORMATTXT
FILE SIZE96 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3812088

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Robust Demographic Inference from Genomic and SNP Data By Laurent Excoffier, Isabelle Dupanloup, Emilia Huerta-Sánchez, Vitor C. Sousa, Matthieu Foll, Joshua M. Akey Published in: PLoS Genetics DOI: 10.1371/journal.pgen.1003905 ABSTRACT We introduce a flexible and robust simulation-based framework to infer demographic parameters from the site frequency spectrum (SFS) computed on large genomic datasets. We show that our composite-likelihood approach allows one to study evolutionary models of arbitrary complexity, which cannot be tackled by other current likelihood-based methods. For simple scenarios, our approach compares favorably in terms of accuracy and speed with , the current reference in the field, while showing better convergence properties for complex models. We first apply our methodology to non-coding genomic SNP data from four human populations. To infer their demographic history, we compare neutral evolutionary models of increasing complexity, including unsampled populations. We further show the versatility of our framework by extending it to the inference of demographic parameters from SNP chips with known ascertainment, such as that recently released by Affymetrix to study human origins. Whereas previous ways of handling ascertained SNPs were either restricted to a single population or only allowed the inference of divergence time between a pair of populations, our framework can correctly infer parameters of more complex models including the divergence of several populations, bottlenecks and migration. We apply this approach to the reconstruction of African demography using two distinct ascertained human SNP panels studied under two evolutionary models. The two SNP panels lead to globally very similar estimates and confidence intervals, and suggest an ancient divergence (>110 Ky) between Yoruba and San populations. Our methodology appears well suited to the study of complex scenarios from large genomic data sets. INTRODUCTION Reconstructing the past history of a given species is important not only for its own sake, but for disentangling demographic from selective effects [1], [2]. Demography is indeed often estimated on a set of markers and the best neutral model is used as a null for evidencing markers under selection [3], [4] or for finding global patterns of selection across the genome [e.g. 5]. Various methods have been proposed to estimate demography from genetic data, including full-likelihood methods [6]–[9], summary-statistics likelihood based methods [10], [11], or different flavours of Approximate Bayesian Computation [12]–[16]. With some exceptions, these methods are relatively slow and do not scale up very well with new genomic data, as computation time increases with the number of loci. In contrast, recently developed composite-likelihood methods based on the site frequency spectrum [SFS, 17] have computing times that do not depend on the amount of available genomic data [18]–[21], and several approaches have been proposed to estimate demographic parameters from the SFS [e.g. 11], [17], [20], [21]–[24]. Among these latter methods, the most widely used is [21], which estimates the expected joint site frequency spectrum for an arbitrary set of parameters by a diffusion approach. Whereas the estimation of the expected SFS is relatively fast, the optimization of the parameters is still time-consuming, which prevents to tackle models with more than three populations at the same time. While some methods can extract demographic information from single whole-genomes per population [25], [26], SFS-based methods, when applied to multiple individuals, do not require whole genome data because correct estimates of the SFS can be obtained from a few Mb [21]. However, with few exceptions [11], the accuracy of SFS-based methods has not been properly assessed, and their ability to infer demographic parameters has been questioned [27]. One advantage of SFS-based inference methods is that they can handle large next generation sequencing (NGS) data sets [28]–[30]. However, the computation of the SFS from NGS data is not always trivial. An empirical Bayes approach has been proposed to estimate the joint 2D SFS from low coverage data [31] and an unbiased maximum likelihood approach has been developed to recover the SFS for a single population [32]. SFS obtained from low-coverage genomic data often show a deficit of rare alleles because a given allele needs to be observed in several individuals to exclude read errors [28], [33]. These missing low frequency variants can lead to imprecisions and biases in population genetic inferences [34]. Several approaches have been proposed to correct for this bias [32], [35], either during the process of genotype calling itself [e.g. 31], [36], [37] or later by applying quality filters on called genotypes [e.g. 38]. Gravel et al. [28] have also proposed to predict the SFS from low-coverage data by using an overlapping subset of high quality data to derive a generalized correction of the SFS. It appears likely that SFS estimation will improve with higher coverage NGS data, and that such data will become increasingly available and used in the near future. As an alternative to deep sequencing, one could use information from a few tens of thousands SNP scattered over the whole genome to make demographic inference, but most SNP chips have complex and often unknown ascertainment schemes that bias the SFS if not properly taken into account [39]–[41]. However, a new SNP chip has recently been introduced [42], [43], which implements a known and simple ascertainment scheme where SNPs are selected at random from sites that are heterozygous in a single individual of a given population. Whereas this ascertainment scheme has no major effect on statistics designed to infer admixture [42], it biases the site frequency spectrum [44], [45] and thus potentially alters the estimation of other parameters. Using simple combinatorics, the SFS can be unbiased [44] in a single population, and this strategy could be extended to unbias joint SFS under complex models involving more populations. A diffusion approach has been recently proposed to estimate divergence times between two populations based on the fraction of SNPs having occurred recently in the ascertained population [45], but this approach is currently restricted to the sole estimation of divergence time and cannot be applied if gene flow occurred between populations. In this paper, we introduce a flexible and robust way to estimate demographic parameters from the SFS inferred from sequence or SNP chip data that we implemented in the fastsimcoal2 software. Our method is based on Nielsen's approach [17], which estimates the expected SFS from simulations under any demographic model. We compare the performance of this approach to [21] under a variety of evolutionary models with simulated data, and we show that it can successfully handle models including more than three populations. We also show how this approach can be extended to deal with ascertained SNP panels by explicitly modelling the ascertainment bias and computing likelihoods based on expected ascertained SFSs. We first apply our method to a large human genomic data set from which we estimate the demography of four populations, and then to two separate Affymetrix ascertained SNP panels [43] from which we estimate the demography of two African populations. RESULTS Comparison between fastsimcoal2 and We performed parameter estimations for 10 data sets generated under each of the 3 evolutionary scenarios shown in Figures 1A–1C. We took two approaches for estimating demography: our new approach based on a composite multinomial likelihood where the expected SFS is obtained using coalescent simulations and [21], which computes a composite Poisson likelihood where the expected SFS is obtained by a diffusion approximation.

Read the complete text