LAVA: a method for identifying local and global adaptation in structured populations

Demonstrating that local adaptation drives phenotypic divergence in quantitative traits requires distinguishing selection from neutral differentiation. Existing methods for detecting selection on quantitative traits, particularly QST - FST comparisons, rely on simplified assumptions about population structure. These methods assume equal relatedness among all subpopulations, which is not necessarily true in a real metapopulation. When assumptions are violated, it has been shown that QST - FST leads to elevated false positive rates. Here we present LAVA, an R implementation of the Log[A]V test: a method for detecting signatures of local and global adaptation in spatially structured populations. LAVA compares two estimates of ancestral additive genetic variance derived from different levels of the demographic history of the metapopulation. Under neutrality, these two estimates should be equal as both reflect the same ancestral additive genetic variance. LAVA additionally allows for the incorporation of environmental covariates to test specific hypotheses about ecological drivers of divergence. LAVA uses a Bayesian linear mixed-effect framework to model population structure through relatedness matrices. We show through extensive simulations across different population structures and selective scenarios how LAVA maintains proper calibration while maintaining or exceeding the power of alternative methods.