The Confounding Effect of Population Structure on Bayesian Skyline Plot Inferences of Demographic History

Abstract

Many coalescent-based methods aiming to infer the demographic history of populations assume a single, isolated and panmictic population (i.e. a Wright-Fisher model). While this assumption may be reasonable under many conditions, several recent studies have shown that the results can be misleading when it is violated. Among the most widely applied demographic inference methods are Bayesian skyline plots (BSPs), which are used across a range of biological fields. Violations of the panmixia assumption are to be expected in many biological systems, but the consequences for skyline plot inferences have so far not been addressed and quantified. We simulated DNA sequence data under a variety of scenarios involving structured populations with variable levels of gene flow and analysed them using BSPs as implemented in the software package BEAST. Results revealed that BSPs can show false signals of population decline under biologically plausible combinations of population structure and sampling strategy, suggesting that the interpretation of several previous studies may need to be re-evaluated. We found that a balanced sampling strategy whereby samples are distributed on several populations provides the best scheme for inferring demographic change over a typical time scale. Analyses of data from a structured African buffalo population demonstrate how BSP results can be strengthened by simulations. We recommend that sample selection should be carefully considered in relation to population structure previous to BSP analyses, and that alternative scenarios should be evaluated when interpreting signals of population size change.

A separate lowstand lake at the northern edge of Lake Tanganyika? Evidence from phylogeographic patterns in the cichlid genus Tropheus

“…Whether the recent population size decline (over last few hundreds of years) apparent in most mitochondrial lineages is a true signal potentially correlated with increasing human population sizes along the lake shore or simply a methodological artifact (Chikhi et al, 2010;Heller et al, 2013) remains unclear. It is also interesting to note that the TCS-1 haplotypes assigned to the pink haplotype cluster are not exclusive to the very northern part of the lake.…”

The devil is in the details: the effect of population structure on demographic inference

“…The risk of such spuriously inferred population size changes has been reported previously (for example, Leblois et al, 2006;Stadler et al, 2009;Heller et al, 2013;Bosse et al, 2014) and the developers of the PSMC method also pointed out that if a population splits in half and later on merges again, even in the absence of demographic changes, an increase in effective population size can be observed (Li and Durbin, 2011). Heller et al (2013) showed how deviations from the random mating model affect demographic inferences using Bayesian skyline plots (BSP), a standard tool used for analysing mitochondrial DNA sequences.…”

“…However, developments in sequencing technologies over the last decade have enabled researchers to produce data covering most, if not all, of the complete genome of a species relatively cheaply (for example, Groenen et al, 2012;Zhan et al, 2013), and estimate important evolutionary parameters using as few as a single genome, for example, the trend in effective population size over time (for example, Li and Durbin, 2011;Schiffels and Durbin, 2014). However, such methodologies assume that the data used for the analysis represent that of an unstructured population, and as it has been shown previously, deviations from that model can inflate estimates of the effective population size (Leblois et al, 2006;Heller et al, 2013;Bosse et al, 2014).…”

“…Heller et al (2013) showed how deviations from the random mating model affect demographic inferences using Bayesian skyline plots (BSP), a standard tool used for analysing mitochondrial DNA sequences. Using sampling schemes where all haploid samples derived from the same subpopulation or from more than one subpopulation, Heller et al demonstrated that for simulations with a stable population size the various sampling regimes mostly resulted in inferred population bottlenecks (Figure 1 in Heller et al, 2013).…”

“…Heller et al (2013) showed how deviations from the random mating model affect demographic inferences using Bayesian skyline plots (BSP), a standard tool used for analysing mitochondrial DNA sequences. Using sampling schemes where all haploid samples derived from the same subpopulation or from more than one subpopulation, Heller et al demonstrated that for simulations with a stable population size the various sampling regimes mostly resulted in inferred population bottlenecks (Figure 1 in Heller et al, 2013). In addition, if simulations included a population expansion or a bottleneck, the former could not be identified if the samples derived form the same subpopulation, and the latter could not be detected if an even amount of samples from each subpopulation was used for the analysis (Figure 2 in Heller et al, 2013).…”