Limits...
The confounding effect of population structure on Bayesian skyline plot inferences of demographic history.

Heller R, Chikhi L, Siegismund HR - PLoS ONE (2013)

Bottom Line: 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.

View Article: PubMed Central - PubMed

Affiliation: Instituto Gulbenkian de Ciência, Oeiras, Portugal. rheller@bio.ku.dk

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.

Show MeSH

Related in: MedlinePlus

The structure effect in a 40-deme constant-size island model.For each scenario, 100 replicate data sets were generated and analysed with EBSPs. Light blue lines represent the median inferred female effective population size through time from each replicate. Time is measured in kya or thousands of years ago and is based on a molecular clock for buffalo D-loop sequences. Bold black lines represent the simulated size of the structured population (500 females * 40 demes = 20,000 females). Insert into each panel is a histogram of PSC values (on x-axis; see main text) across replicates. Dashed lines show the prior distribution for PSC. The y-axis in the insert histograms marks the frequency of occurrence in each PSC bin out of 100 replicates.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3646956&req=5

pone-0062992-g001: The structure effect in a 40-deme constant-size island model.For each scenario, 100 replicate data sets were generated and analysed with EBSPs. Light blue lines represent the median inferred female effective population size through time from each replicate. Time is measured in kya or thousands of years ago and is based on a molecular clock for buffalo D-loop sequences. Bold black lines represent the simulated size of the structured population (500 females * 40 demes = 20,000 females). Insert into each panel is a histogram of PSC values (on x-axis; see main text) across replicates. Dashed lines show the prior distribution for PSC. The y-axis in the insert histograms marks the frequency of occurrence in each PSC bin out of 100 replicates.

Mentions: Our first series of simulations with no demographic change demonstrates that population structure can mimic population size changes in the absence of any such change (Fig. 1). The probability of such a misinterpretation depends on the interplay between the sampling strategy and the level of gene flow. Locally sampled scenarios under all levels of migration as well as pooled samples from the lowest migration class showed a clear trend towards declining EBSPs. The posterior distribution of the population size change (PSC) parameter confirmed this qualitative observation, as the above scenarios showed a strong tendency towards overestimating PSC (Fig. 1 insert panels). The measures of coverage and mean relative departure (MRD) corroborated the visual inspection of EBSPs and the PSC distributions, showing that the inferred population size at the present was notoriously lower than the simulated one under local sampling (Table 1). Other notable observations include a consistently higher inferred population size than the simulated one in the older parts of the EBSPs under scattered and pooled sampling for the two lowest migration classes (though less pronounced for the pooled strategy). This discrepancy between effective and actual (i.e total number of individuals) population size at low migration rates under scattered sampling is in fact expected as shown in Wakeley’s Eq. (6.18) [26]. The MRD of the inferred from the simulated population sizes is close to the expected magnitude of this effect (expected MRD from Wakeley’s Eq. (6.18): 4.00, 0.40 and 0.04 at Nm = 0.125, 1.25 and 12.5, respectively, to be compared with our calculated values of 4.33, 0.54 and 0.08; Table 1), showing the ability of EBSPs to correctly infer the structured effective population size in the collecting phase.


The confounding effect of population structure on Bayesian skyline plot inferences of demographic history.

Heller R, Chikhi L, Siegismund HR - PLoS ONE (2013)

The structure effect in a 40-deme constant-size island model.For each scenario, 100 replicate data sets were generated and analysed with EBSPs. Light blue lines represent the median inferred female effective population size through time from each replicate. Time is measured in kya or thousands of years ago and is based on a molecular clock for buffalo D-loop sequences. Bold black lines represent the simulated size of the structured population (500 females * 40 demes = 20,000 females). Insert into each panel is a histogram of PSC values (on x-axis; see main text) across replicates. Dashed lines show the prior distribution for PSC. The y-axis in the insert histograms marks the frequency of occurrence in each PSC bin out of 100 replicates.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3646956&req=5

pone-0062992-g001: The structure effect in a 40-deme constant-size island model.For each scenario, 100 replicate data sets were generated and analysed with EBSPs. Light blue lines represent the median inferred female effective population size through time from each replicate. Time is measured in kya or thousands of years ago and is based on a molecular clock for buffalo D-loop sequences. Bold black lines represent the simulated size of the structured population (500 females * 40 demes = 20,000 females). Insert into each panel is a histogram of PSC values (on x-axis; see main text) across replicates. Dashed lines show the prior distribution for PSC. The y-axis in the insert histograms marks the frequency of occurrence in each PSC bin out of 100 replicates.
Mentions: Our first series of simulations with no demographic change demonstrates that population structure can mimic population size changes in the absence of any such change (Fig. 1). The probability of such a misinterpretation depends on the interplay between the sampling strategy and the level of gene flow. Locally sampled scenarios under all levels of migration as well as pooled samples from the lowest migration class showed a clear trend towards declining EBSPs. The posterior distribution of the population size change (PSC) parameter confirmed this qualitative observation, as the above scenarios showed a strong tendency towards overestimating PSC (Fig. 1 insert panels). The measures of coverage and mean relative departure (MRD) corroborated the visual inspection of EBSPs and the PSC distributions, showing that the inferred population size at the present was notoriously lower than the simulated one under local sampling (Table 1). Other notable observations include a consistently higher inferred population size than the simulated one in the older parts of the EBSPs under scattered and pooled sampling for the two lowest migration classes (though less pronounced for the pooled strategy). This discrepancy between effective and actual (i.e total number of individuals) population size at low migration rates under scattered sampling is in fact expected as shown in Wakeley’s Eq. (6.18) [26]. The MRD of the inferred from the simulated population sizes is close to the expected magnitude of this effect (expected MRD from Wakeley’s Eq. (6.18): 4.00, 0.40 and 0.04 at Nm = 0.125, 1.25 and 12.5, respectively, to be compared with our calculated values of 4.33, 0.54 and 0.08; Table 1), showing the ability of EBSPs to correctly infer the structured effective population size in the collecting phase.

Bottom Line: 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.

View Article: PubMed Central - PubMed

Affiliation: Instituto Gulbenkian de Ciência, Oeiras, Portugal. rheller@bio.ku.dk

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.

Show MeSH
Related in: MedlinePlus