The aggregate site frequency spectrum for comparative population genomic inference

Abstract

Understanding how assemblages of species responded to past climate change is a central goal of comparative phylogeography and comparative population genomics, an endeavor that has increasing potential to integrate with community ecology. New sequencing technology now provides the potential to perform complex demographic inference at unprecedented resolution across assemblages of non-model species. To this end, we introduce the aggregate site frequency spectrum (aSFS), an expansion of the site frequency spectrum to use single nucleotide polymorphism (SNP) datasets collected from multiple, co-distributed species for assemblage-level demographic inference. We describe how the aSFS is constructed over an arbitrary number of independent population samples and then demonstrate how the aSFS can differentiate various multi-species demographic histories under a wide range of sampling configurations while allowing effective population sizes and expansion magnitudes to vary independently. We subsequently couple the aSFS with a hierarchical approximate Bayesian computation (hABC) framework to estimate degree of temporal synchronicity in expansion times across taxa, including an empirical demonstration with a dataset consisting of five populations of the threespine stickleback (Gasterosteus aculeatus). Corroborating what is generally understood about the recent post-glacial origins of these populations, the joint aSFS/hABC analysis strongly suggests that the stickleback data are most consistent with synchronous expansion after the Last Glacial Maximum (posterior probability = 0.99). The aSFS will have general application for multi-level statistical frameworks to test models involving assemblages and/or communities and as large-scale SNP data from non-model species become routine, the aSFS expands the potential for powerful next-generation comparative population genomic inference.

Inferring responses to climate dynamics from historical demography in neotropical forest lizards

“…Extending beyond regionspecific inferences of population size change within each species, we applied hierarchical demographic models to combine data across regions and species and test for assemblage-wide synchrony in population shifts. To implement this multiregion, multitaxon framework, we used the aggregate site frequency spectrum (aSFS) to summarize patterns of genetic variation across groups and coupled it with coalescent simulations and a hierarchical approximate Bayesian computation (hABC) approach that allows species-specific parameters to vary independently (22). In an analysis that combined data from both A. punctatus and A. ortonii in all regions (western Amazonia, eastern Amazonia, and northern Atlantic Forest), we recovered a signal of largely asynchronous demographic change [dispersal index, the variance/mean of the time of population size change (DI) = 76,549; proportion of groups in a synchronous pulse (ζ) = 0.50; Dataset S5 and Fig.…”

“…For this analysis, all three species in the northern Atlantic Forest were down-projected to four diploid individuals. In each analysis, we combined the empirical (observed) SFS of the spatial groups under consideration into an aSFS (22).…”

“…We then used an hABC procedure, which compares the observed aSFS with aSFS simulated under a hierarchical demographic model with varying degrees of synchronicity, to estimate (i) the proportion of groups in synchrony, ζ; (ii) the variance/mean of demographic change timing across spatial groups (with lower values signaling greater synchronicity), DI; and (iii) the timing of synchronous population size change (22). We performed 1 million coalescent simulations of the aSFS using fastsimcoal2, implementing the same prior distributions as the demographic syndrome analysis; the exception was the prior for e in the expander populations, set to ∼U(0.02, 0.2), and for current effective population size, set to ∼U(500 k, 2 million) for expanders and ∼U(50 k, 250 k) for contractors, because single-population analyses often had L-shaped posterior distributions for these parameters.…”

“…Our approach tests for temporal congruence of demographic trends, using coalescent simulations under a hierarchical demographic model in combination with approximate Bayesian computation (ABC) to test for synchronous shifts in population sizes (22). We examine the timing and magnitude of population changes in different forest regions occupied by a single species as well as codistributed species within the same region.…”

“…Methods that build upon coalescent theory to account for historical heterogeneity across taxa under a single statistical framework (21,22) provide increased power to test concerted demographic responses at the level of ecological assemblages. These highly flexible approaches have profited from novel bench and analytical methods that build on genomic or reduced genome data, resulting in improved estimates of demographic parameters and histories (23)(24)(25).…”