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

Abstract

Population genetics model based Bayesian methods have been proposed and widely applied to making unsupervised inference of population structure from a sample of multilocus genotypes. Usually they provide good estimates of the ancestry (or population membership) of sampled individuals by clustering them probabilistically or proportionally into (anonymous) populations. However, they have difficulties in accurately estimating the number of populations (K) represented by the sampled individuals. This study proposed a new ad hoc estimator of K, calculable from the output of a population clustering program such as STRUCTURE or ADMIXTURE. The new criterion, called parsimony index (PI), aims to identify the number of populations (K) which yields consistently the minimal admixture estimates of sampled individuals. Extensive simulated and empirical data were used to compare the accuracy of PI and two popular K estimators based on Pr[X|K] (i.e., the probability of genotype data X given K) and ΔK (i.e., the rate of change of the probability of data as a function of K) calculated from STRUCTURE outputs, and the accuracy of PI and the cross‐validation method calculated from ADMIXTURE outputs. It was shown that PI was more accurate than the other methods consistently in various population structure (e.g., hierarchical island model, different extents of differentiation) and sampling (e.g., unbalanced sample sizes, different marker information contents) scenarios. The ΔK method was more accurate than the Pr[X|K] method only for hierarchically structured or highly inbred populations, and the opposite was true in the other scenarios. The PI method was implemented in a computer program, KFinder, which can be run on all major computer platforms.

Conservation genetics of the pond bat (Myotis dasycneme) with special focus on the populations in northwestern Germany and in Jutland, Denmark

“…Consequently, STRUCTURE analysis together with all the data analysis based on microsatellite markers were performed based on the remaining seven loci and the aggregated Danish samples. The three different approaches implemented in KFinder (Wang, ) to identify the individuals' ancestry from the STRUCTURE (Pritchard et al, ) results, all returned two populations (best K = 2) as the most probable structure (Table ; Figure ). The structure found was not clear, but inspecting the output files suggested that Denmark and Germany belonged to one population and Russia to another population (data not shown).…”

“… Three different approaches were applied to interpret the best number of clusters (best K ) obtained from STRUCTURE: the mean likelihood Pr[X|K] that maximize K (Pritchard et al, ), ΔK (the largest rate of change of the log probability given the data) (Evanno et al, ; STRUCTURE HARVESTER (Earl & VonHoldt, )), and the Parsimony Index (KFinder, Wang, ), identifying the K , which repeatedly returns the minimal mean admixture of the sample. K = number of assumed clusters.…”

“…The clusters of individuals forming the number of populations with the highest likelihood were assigned to sampling locations. As it might be difficult to infer the number of clusters represented due to an effect of isolation by distance (IBD) and extensive admixture (Falush, Stephens, & Pritchard, ; Pritchard et al, ), KFinder (Wang, ) was applied to infer the number of clusters. KFinder includes three different criterions for assessing the most likely number of populations represented by the sample of individuals.…”

Comparative landscape genetics reveals the evolution of viviparity reduces genetic connectivity in fire salamanders

“…the K value that maximises the mean logarithmic posterior probability (ln[X|K]; Pritchard et al, 2000); (b) the rate of change of ln [X|K] among K values (ΔK; Evanno, Regnaut, & Goudet, 2005); and (c) the parsimony index (PI) parameter, which identifies the K that yields the most consistent and minimal average admixture (Wang, 2019).…”

“…Extensive simulations showed PI more accurately infers the correct number of genetic clusters under a variety of scenarios, including unbalanced sampling, low numbers of loci, reduced genetic divergence, and inbreeding (Wang, 2019).…”