Landscape modelling of gene flow: improved power using conditional genetic distance derived from the topology of population networks

Abstract

Landscape genetics is a burgeoning field of interest that focuses on how site-specific factors influence the distribution of genetic variation and the genetic connectivity of individuals and populations. In this manuscript, we focus on two methodological extensions for landscape genetic analyses: the use of conditional genetic distance (cGD) derived from population networks and the utility of extracting potentially confounding effects caused by correlations between phylogeographic history and contemporary ecological factors. Individual-based simulations show that when describing the spatial distribution of genetic variation, cGD consistently outperforms the traditional genetic distance measure of linearized F(ST) under both 1- and 2-dimensional stepping stone models and Cavalli-Sforza and Edward's chord distance D(c) in 1-dimensional landscapes. To show how to identify and extract the effects of phylogeographic history prior to embarking on landscape genetic analyses, we use nuclear genotypic data from the Sonoran desert succulent Euphorbia lomelii (Euphrobiaceae), for which a detailed phylogeographic history has previously been determined. For E. lomelii, removing the effect of phylogeographic history significantly influences our ability to infer both the identity and the relative importance of spatial and bio-climatic variables in subsequent landscape genetic analyses. We close by discussing the utility of cGD in landscape genetic analyses.

Different patterns of colonization of Oxalis alpina in the Sky Islands of the Sonoran desert via pollen and seed flow

“…Therefore, longer edges indicate lower genetic covariance between populations. Populations that are not connected indicate the absence of migration, and the presence of subgraphs (a smaller network within a large network) indicates that a population or group of populations maintain a weak or null genetic connection (Dyer, 2007; Dyer & Nason, 2004; Dyer, Nason, & Garrick, 2010). In addition, we tested for a correlation between the matrices of genetic distances (Nei, 1972) and the midmorph frequencies using a Mantel test (ADE4 package Dray & Dufour, 2007, in R).…”

Network modularity reveals critical scales for connectivity in ecology and evolution

“…For bullfrogs, genetic distances are probably influenced by both historical and contemporary ecological processes 41 ; in this instance, modularity may be confounding limited gene flow between patches with the historical process of secondary contact between allopatric lineages. Bullfrogs in Ontario represent two distinct postglacial lineages, whose distribution is very similar to that represented by the observed modules 42 after the effects of distance are removed (Fig.…”

Phylogeography unplugged: comparative surveys in the genomic era

“…Population graphs connect the population pairs exchanging migrants in the network by edges whose length is inversely proportional to the genetic covariance between the populations (Dyer and Nason 2004). Because conditional genetic distance (cGD) in a population graph is based upon the differences in genetic covariation associated with both direct and indirect gene flow among populations, cGD should be more sensitive than conventional pairwise F ST and better suited for landscape genetics (Dyer et al 2010). Another promising avenue incorporating genetic data sets for conservation planning would be Bayesian Belief Network analysis (BBNs), which can overcome variable sampling coverage across data sets to build a single probability distribution (e.g., Kininmonth et al 2010a,b).…”

“…Therefore, HABC approaches are expected to provide a bridge linking comparative phylogeography with community ecology (Hickerson and Meyer 2008). On another front, population graph approaches can build congruence networks based on two or more spatial genetic data sets (Dyer et al 2010). Population graphs connect the population pairs exchanging migrants in the network by edges whose length is inversely proportional to the genetic covariance between the populations (Dyer and Nason 2004).…”