Sampling strategy optimization to increase statistical power in landscape genomics: a simulation-based approach

An increasing number of studies use landscape genomics to investigate local adaptation in wild and domestic populations. This method consists of simultaneously characterizing genetic and environmental variability across a landscape in order to depict how genotype frequencies are shaped by adaptive forces. The implementation of this approach requires the sampling phase to consider the complexity of environmental settings and the burden of logistic constraints. These aspects can have a major impact on reproducibility and costs but they are often underestimated in the literature dedicated to sampling strategies. In this study, we compute simulated genomic datasets on a real environmental setting and distinct demographic scenarios to reiterate landscape genomics experiments with distinct sampling strategies. The latter differ by design approaches (aiming at enhancing environmental and/or geographic representativeness), number of sampling locations and sample sizes. We then propose an analytical framework to evaluate the effect of these elements on statistical power and false discovery rates of the analysis. Our results highlight the importance of sample size in designing landscape genomic studies, where a minimal threshold of 200 individuals is required to detect at least part of the true positives. Furthermore, our results show that sampling at less than 10 sites clearly impairs statistical power, but also that the optimal number of sampling locations is not necessarily the highest possible. Last, we show that the designs that account for environmental representativeness result in the highest statistical power. Overall, this study provides valuable guidelines for optimizing sampling strategies in landscape genomics experiments.