Reliably discriminating stock structure with genetic markers: Mixture models with robust and fast computation

Abstract

Delineating naturally occurring and self-sustaining subpopulations (stocks) of a species is an important task, especially for species harvested from the wild. Despite its central importance to natural resource management, analytical methods used to delineate stocks are often, and increasingly, borrowed from superficially similar analytical tasks in human genetics even though models specifically for stock identification have been previously developed. Unfortunately, the analytical tasks in resource management and human genetics are not identical-questions about humans are typically aimed at inferring ancestry (often referred to as "admixture") rather than breeding stocks. In this article, we argue, and show through simulation experiments and an analysis of yellowfin tuna data, that ancestral analysis methods are not always appropriate for stock delineation. In this work, we advocate a variant of a previously introduced and simpler model that identifies stocks directly. We also highlight that the computational aspects of the analysis, irrespective of the model, are difficult. We introduce some alternative computational methods and quantitatively compare these methods to each other and to established methods. We also present a method for quantifying uncertainty in model parameters and in assignment probabilities. In doing so, we demonstrate that point estimates can be misleading. One of the computational strategies presented here, based on an expectation-maximization algorithm with judiciously chosen starting values, is robust and has a modest computational cost.

Population differentiation from environmental DNA: Investigating the potential of haplotype presence/absence‐based analysis of molecular variance

“…The graph "All individuals" represents the proportion of significant results obtained when performing an AMOVA using all the haplotypes generated in each population used to analyze eDNA data so that this tool can be rigorously use for population genetics. Here, our study focused on a hypothesis-testing method, but a similar approach could be used to assess the power of hypothesis-generating algorithms such as Bayesian clustering analysis of population structure (Corander & Tang, 2007), AMOVA-based Kmeans clustering (Meirmans, 2012) or the power of models for stock identification (Foster et al, 2018) with eDNA data.…”

Spatiotemporal clustering using Gaussian processes embedded in a mixture model

“…Moreover, there are no spatial replicates from exactly same sampling sites even though sites at different times may be located close to each other. Further complexity stems from the fact that a mixture model's likelihood and posterior densities are known to be "bumpy" with many local maxima (Foster et al, 2018;McLachlan & Peel, 2000). In order to make analyses feasible, we propose to use approximate Bayesian inference methods in three increasing levels of accuracy and computation time.…”

One panel to rule them all: DArTcap genotyping for population structure, historical demography, and kinship analyses, and its application to a threatened shark

“…Finally, the R packages Radiator (Gosselin, Lamothe, Devloo‐Delva, & Grewe, 2020) and OutFLANK (Whitlock & Lotterhos, 2015) were used to detect and remove sex‐linked and outlier loci respectively. Since unequal sample sizes could introduce bias in clustering algorithms (Foster et al., 2018), filtering was applied to the full data set and a subsampled data set including only 30 randomly chosen sharks from Van Diemen Gulf (VDG) was selected so that putative populations would roughly be of equal size (Supporting Information S1, section 6).…”