Relating processes to patterns of genetic variation across landscapes

Contemporary patterns of genetic variation can be incorporated into theoretical models with the goal of understanding the evolutionary processes that led to these patterns. Neutral processes allow explicit predictions to be made, such that a comparison across loci permits inferences about mutation, migration and drift. Because migration and drift act evenly on all nuclear loci, we can infer that those loci which show discordant patterns of variation may be experiencing either contrasting mutation rates, or locus-specific natural selection. However, in changing environments, populations are frequently out of equilibrium with patterns of migration and drift. This can induce a high locus-to-locus heterogeneity in patterns of variation, mimicking the action of natural selection. Coalescent modeling indicates that even highly contrasting patterns of variation can be consistent with neutral evolution at all loci. The same reasoning can be applied to quantitative traits, provided that genetic variation in traits is explicitly measured. Generally, traits exhibit greater population divergence than do putatively neutral genetic markers, an observation consistent with locally adaptive selection acting on traits. However, the correlation of marker divergence with trait divergence is not strong enough to be predictive. The individual loci underlying quantitative traits (QTLs) are the hardest to study directly in natural populations, and simulations suggest that such loci will exhibit patterns very similar to neutral marker loci, in spite of strong selection on the traits they code for. This can occur because spatially heterogeneous selection imposes a covariance of allele frequencies across populations, so that traits diverge to a greater extent than the allele frequencies at the corresponding QTLs. This holds out the possibility of using simple genetic markers to draw inferences about the distribution and amount of allelic variation (adaptive potential) at QTLs, but such conjecture remains to be empirically confirmed.