Genetic offset and vulnerability modelling: misinterpretations of results and violations of evolutionary principles 

Abstract

Genetic offset models have become a popular component of the landscape genetics toolbox, with over 150 peer-reviewed publications applying these models to plant and animal systems. Genetic offset models are most frequently performed following the identification of putatively adaptive alleles from genotype-environment association analyses in natural populations of non-model organisms. These models allow the researcher to make predictions about the likely vulnerability of species populations to climate change, by estimating the extent of genetic change needed (i.e., genetic offset) to maintain ‘optimal’ allele frequencies and population fitness under future climate change scenarios. However, a number of critical assumptions and knowledge gaps are often overlooked when undertaking these analyses, undermining their reliability and usefulness for making genetically informed management decisions. In this commentary, we describe seven assumptions in detail and discuss a range of common violations that can lead to misleading outcomes. While genetic offset models may have a place for predicting the vulnerability of natural populations of non-model organisms in the future, we argue that current applications are likely to be problematic and risk both wasting resources and misleading conservation management.

Evaluating genomic offset predictions in a forest tree with high population genetic structure

“…Towards that end, developing a more solid theoretical framework around the GO concept appears necessary. Progress has already been made in defining a quantitative theory for GO statistics (Gain et al 2023) and identifying limitations of the GO approach (Hoffmann et al 2021, Rellstab et al 2021, Ahrens et al 2023. Empirical validation based on multiple data sources (e.g., natural populations vs common gardens, multi-environment common gardens, multiple traits, multiple species, etc) will also be needed before the GO concept can be confidently applied in conservation and management strategies as a metric of maladaptation.…”

“…Importantly, we must remain cautious when interpreting GO predictions, especially when the aim is to use them to inform conservation or management strategies. The limits of the GO approach have already been discussed in depth (Hoffmann et al 2021, Rellstab et al 2021, Ahrens et al 2023), so we will mention here just a few arguments that we consider should be given particular attention in maritime pine, and more generally in forest trees. First, climate change exposure is integrated within the GO approach using distances between long-term means of past and future climatic conditions.…”

“…In this context, it is crucial to emphasize that GO is not informative about the capacity of populations to adapt, i.e., to evolve towards the new fitness optima of future climates, and therefore captures only part of the vulnerability of populations to climate change (IPCC 2007, Foden et al 2019). Moreover, the GO concept relies on strong assumptions that undermine the robustness of its predictions (Rellstab et al 2021, Ahrens et al 2023). One key assumption is that populations are currently optimally adapted to their local environment, which is often violated.…”

Genomics‐informed conservation units reveal spatial variation in climate vulnerability in a migratory bird

“…Higher genomic offset indicates that populations will need to change allele frequencies more to maintain current gene–environment relationships in future conditions. Though some caution should be used when evaluating genomic offset metrics due to the assumptions of the model (for a thorough discussion of assumptions and limitations, see Ahrens et al., 2023; Capblancq et al., 2020; DeSaix et al., 2022; Rellstab, 2021), here we use genomic offset to identify which AUs may be the most at risk of climate‐related vulnerability. Using the adaptive loci found with LFMM and RDA as our response, we ran gradient forest (Ellis et al., 2012) and used the reduced environmental variable set as predictors to generate a model of allele frequency turnover across the breeding range.…”

The genetic consequences of population marginality: a case study in maritime pine

“…Estimation of the genomic offset is becoming a popular approach to assess population vulnerability in the face of climate change, but it is not free of pitfalls (see Ahrens et al, 2023;Rellstab et al, 2021;Lind et al, 2023;Archambeau et al, forthcoming). Moreover, the genomic offset can gauge for some components of population vulnerability (i.e., exposure and sensitivity; Estoque et al, 2023;IPCC, 2007) but not others (notably the adaptive capacity of populations; Fitzpatrick et al, 2021;Archambeau et al, forthcoming).…”