Data from: Seeing the Forest for the trees: Assessing genetic offset predictions from Gradient Forest

Gradient Forest (GF) is a machine learning algorithm designed to analyze
spatial patterns of biodiversity as a function of environmental gradients.
An offset measure between the GF predicted environmental association of
adapted alleles and a new environment (GF Offset), is increasingly being
used to predict the loss of environmentally adapted alleles under rapid
environmental change, but remains mostly untested for this purpose. Here
we explore the robustness of GF Offset to assumption violations, and its
relationship to measures of fitness, using SLiM simulations with explicit
genome architecture and a spatial metapopulation. We evaluate measures of
GF Offset in: (1) a neutral model with no environmental adaptation; (2) a
monogenic “population genetic” model with a single environmentally adapted
locus; and (3) a polygenic “quantitative genetic” model with two adaptive
traits, each adapting to a different environment. We found GF Offset to be
broadly correlated with fitness offsets under both single locus and
polygenic architectures. However, neutral demography, genomic
architecture, and the nature of the adaptive environment can all confound
relationships between GF Offset and fitness. GF Offset is a promising
tool, but it is important to understand its limitations and underlying
assumptions, especially when used in the context of predicting
maladaptation.