Effect of sampling close-related individuals breeding in small container habitats over population genetic estimators: a modeling approach

Background/Question/Methods

Many organisms breed in small, contained environments shared by multiple individuals, including small container-breeding mosquitoes. Because these mosquitoes can be important vectors of human arboviruses, the accurate determination of biological and genetic parameters is fundamental for the understanding of their population biology. Sampling immature individuals from their breeding habitat is often an efficient strategy. However, this approach is sensitive to the inclusion of non-independent samples (i.e., sibling) sharing the same breeding habitat. This is a common consequence of females depositing multiple eggs in a particular habitat. Many studies take preventive measures when collecting samples for genetic studies; frequently, studies restrict the number of samples from an individual habitat to five, despite the lack of studies assessing this practice. In this study, we used a stochastic R script to simulate sets of genetic samples where the effect of identical alleles (siblings) over the estimation of commonly used population genetic estimators is evaluated. Our script allows to set the limit of samples used from each individual sampled container and the number of individual sampling habitat considered. We explore the effects of commonly used approaches to prevent the presence of non-independent samples in the population genetic estimators.

Results/Conclusions

Using a novel script that allows to explore a wide range of situations, our study determines that sampling siblings generate biased estimators. Estimators can be robust when precautions are taken to limit the number of individual samples used per container habitat and diversity exists among sampled individuals in each habitat even if non-independent samples are included in the analyses (data extremes overlap). However, when multiple individual samples are identical (siblings) in an individual container, the estimators are extremely biased (data extremes do not overlap). In some scenarios, complete avoidance of multiple individuals per habitat performs better (data extremes overlapping). We discuss the implications of reducing the sampling size to avoid including close related individuals vs the loss of power. Finally, we propose that simple exploratory analyses may help researchers to consider for an optimal compromise between sampling representativity and bias avoidance.