A fast likelihood solution to the genetic clustering problem

Abstract

The investigation of genetic clusters in natural populations is an ubiquitous problem in a range of fields relying on the analysis of genetic data, such as molecular ecology, conservation biology and microbiology. Typically, genetic clusters are defined as distinct panmictic populations, or parental groups in the context of hybridisation. Two types of methods have been developed for identifying such clusters: model‐based methods, which are usually computer‐intensive but yield results which can be interpreted in the light of an explicit population genetic model, and geometric approaches, which are less interpretable but remarkably faster.Here, we introduce snapclust, a fast maximum‐likelihood solution to the genetic clustering problem, which allies the advantages of both model‐based and geometric approaches. Our method relies on maximising the likelihood of a fixed number of panmictic populations, using a combination of geometric approach and fast likelihood optimisation, using the Expectation‐Maximisation (EM) algorithm. It can be used for assigning genotypes to populations and optionally identify various types of hybrids between two parental populations. Several goodness‐of‐fit statistics can also be used to guide the choice of the retained number of clusters.Using extensive simulations, we show that snapclust performs comparably to current gold standards for genetic clustering as well as hybrid detection, with some advantages for identifying hybrids after several backcrosses, while being orders of magnitude faster than other model‐based methods. We also illustrate how snapclust can be used for identifying the optimal number of clusters, and subsequently assign individuals to various hybrid classes simulated from an empirical microsatellite dataset. snapclust is implemented in the package adegenet for the free software R, and is therefore easily integrated into existing pipelines for genetic data analysis. It can be applied to any kind of co‐dominant markers, and can easily be extended to more complex models including, for instance, varying ploidy levels. Given its flexibility and computer‐efficiency, it provides a useful complement to the existing toolbox for the study of genetic diversity in natural populations.

A graph-based evidence synthesis approach to detecting outbreak clusters: An application to dog rabies

“…For the analysis of the baseline simulations, we systematically varied the cutoffs and reporting rate, as described in S2 Table. Detailed information on the simulation process is available in S1 Text. Following Beugin et al [35], we quantify the ability of our method to correctly identify clusters of cases linked by transmission, through measuring 1) the true positive rate (TPR) or sensitivity, i.e. the proportion of pairs of cases belonging to the same transmission tree who are inferred to be in the same outbreak cluster), 2) the true negative rate (TNR) or specificity, i.e.…”

“…the proportion of pairs of cases belonging to the same transmission tree who are inferred to be in the same outbreak cluster), 2) the true negative rate (TNR) or specificity, i.e. the proportion of pairs of cases not belonging to the same transmission tree who are assigned to different outbreak clusters and 3) the mean between TPR and TNR, which is proportional to the Rand index, a common criterion used to evaluate clustering methods [35,70]. We also compare the estimates of the reproduction number and importation rate to the values used in the simulation.…”

“…sources of data such as time, space, and WGS [29][30][31][32]. Lastly, population genetics has had a long-standing tradition of developing clustering methods [33][34][35], some of which have proved useful for studying pathogen populations [36,37]. Unfortunately, despite these connections to well-developed methodological fields, the integration of multiple data sources (e.g.…”

Population substructure and signals of divergent adaptive selection despite admixture in the sponge Dendrilla antarctica from shallow waters surrounding the Antarctic Peninsula

“…Population structure was assessed using the function snapclust in the R package adegenet (Beugin, Gayet, Gayet, Pontier, Devillard, & Jombart, ), the discriminant analysis of principal components (DAPC) as implemented in the adegenet R package (Jombart, Devillard, & Balloux, ), and structure version 2.3 (Pritchard, Stephens, & Donnelly, ). structure and snapclust may produce similar individual membership probability plots, but they have totally different approaches to the genetic clustering problem: while structure uses a Bayesian approach with Markov chain Monte Carlo (MCMC) method to estimate allele frequencies in each cluster and population memberships for every individual, snapclust is a fast likelihood optimization method combining both model‐based and geometric clustering approaches, which uses the Expectation‐Maximization (EM) algorithm to assign genotypes to populations and detect admixture patterns.…”

A parsimony estimator of the number of populations from a STRUCTURE‐like analysis

“…Corander et al (2003) and Corander, Waldmann, Marttinen, and Sillanpää (2004) implemented a split-and-merge algorithm in their program BAPS to estimate K. Patterson, Price, and Reich (2006) proposed an eigenanalysis method, implemented in SmartPCa software, to estimate K as 1 plus the number of significant eigenvalues explaining the variation of genotype data. Jombart et al (2010) and Beugin, Gayet, Pontier, Devillard, and Jombart (2018) used Akaike information criterion (AIC: Akaike, 1998), Bayesian Information Criterion (BIC: Schwarz, 1978), Kullback Information Criterion (KIC: Cavanaugh, 1999) and their variants to assess the best supported model, and therefore the most likely number of populations. These and other methods were demonstrated to yield good estimates of K in some simple scenarios (e.g., Gao, Bryc, & Bustamante, 2011), but can be highly inaccurate in difficult situations such as many source populations (e.g., K > 10), unbalanced sample sizes (Wang, 2017), hierarchical population structures (Evanno et al, 2005), weak differentiation or low marker information (Gao et al, 2011), and high admixture.…”