Fig 1.
Simulation scenarios and their corresponding geogenetic maps estimated with SpaceMix.
The smaller circles in the simulation scenarios represent unsampled populations. a) the configuration of simulated populations on a simple lattice with spatially homogeneous migration rates (a plot showing the first two Principal Component axes for this simulation is given in a); b) a lattice with a barrier along the center line of longitude, across which migration rates are reduced by a factor of 5; c) a lattice with recent expansion on the eastern margin; d) the maximum a posteriori (MAP) estimate from the posterior distribution of population locations under the scenario in 1a; e) MAP estimate of population locations under the scenario in 1b; f) MAP estimate of population locations under the scenario in 1c.
Fig 2.
Simulation scenarios and SpaceMix inference.
a) a lattice with a recent admixture event between population 1 in the southwest corner and population 30 in the northeast corner, so that population 30 is drawing half of its ancestry from population 1 (a plot showing the first two Principal Component axes for this simulation is given in S2 Fig panel b); b) the estimate of population locations under this scenario; c) the estimate of population locations and their sources of admixture under this scenario. The 95% credible interval on w30 is 0.36–0.40. In panel (c), the width and opacity of the admixture arrows are drawn proportional the admixture proportions.
Fig 3.
An illustration of the form of the admixed covariance.
Following Eq (6), populations i and j are drawing admixture in proportions wi and wj from their respective sources of admixture, i* and j*, and all pairwise spatial covariances (the F’s) are shown. In this cartoon example, population j is drawing more admixture from its source j* than i is from its source i* (i.e., wj > wi).
Fig 4.
Simulation scenarios and inferred population maps for two different admixture scenarios.
Green arrows denote admixture from a source to a target population in the simulation. a) lattice with a barrier and an admixture event (10%) across the barrier to an ‘inland’ population; b) the inferred population map for the scenario in (a), where the admixed population 23 is the only population drawing non-negligible admixture (95% CI: 0.02-0.08); c) lattice with a barrier and an admixture event (40%) across the barrier to a ‘neighbor’ population on the border of the barrier; (d) the inferred population map for the scenario in (c), where the admixed population 18 is the only population drawing non-negligible admixture (95% CI: 0.04–0.14).
Fig 5.
Sampling maps of both empirical systems analyzed.
(a) greenish warbler subspecies distributions of all 22 sampled populations (breeding grounds), consisting of 95 individuals and colored by subspecies [46]; (b) sampling map for human dataset, consisting of 1,490 individuals from 95 population samples [50].
Fig 6.
Inferred maps for warbler populations.
Population labels are colored as in Fig 5a. a) the map inferred with no admixture inference; b) the map inferred with admixture inference.
Fig 7.
Inferred maps for warbler individuals with admixture inference.
Individual labels are colored by subspecies as in Fig 5a. a) map inferred with admixture; b) close-up of all non-nitidus samples in the admixture map.
Fig 8.
Geogenetic map of human samples, inferred without admixture.
a) complete map; b) close-up of Eurasian samples.
Fig 9.
Geogenetic map of human samples, inferred with admixture.
Labels are colored as in as in Fig 5b. Italicized labels denote locations of admixture sources, with opacity proportional to the amount of admixture drawn by the sample. a) complete map; b) close-up of Eurasian samples.
Fig 10.
Admixture proportions (95% CIs) for each human population sample.
Labels are colored as in as in Fig 5b.
Table 1.
List of models that may be specified using SpaceMix, along with the number and identity of free parameters in each.
Table 2.
List of parameters used in the SpaceMix models, along with their descriptions and priors.
is the mean of the pairwise distances between observed locations G(obs).
Fig 11.
Illustrated example of spatial covariance and the effects of admixture.
Lines show the covariances populations A, C, and D would have with population B as a function of B’s location with no admixture, under the parametric form of Eq (2). The colored dots above ‘B’ show the covariances observed with B at that location given that B has 40% admixture from D. There is no spatial configuration that induces unadmixed covariances remotely similar to those observed.