Simulated Disperser Analysis: determining the number of loci required to genetically identify dispersers

Abstract

Empirical genetic datasets used for estimating contemporary dispersal in wild populations and to correctly identify dispersers are rarely tested to determine if they are capable of providing accurate results. Here we test whether a genetic dataset provides sufficient information to accurately identify first-generation dispersers. Using microsatellite data from three wild populations of common starlings (Sturnus vulgaris), we artificially simulated dispersal of a subset of individuals; we term this 'Simulated Disperser Analysis'. We then ran analyses for diminishing numbers of loci, to assess at which point simulated dispersers could no longer be correctly identified. Not surprisingly, the correct identification of dispersers varied significantly depending on the individual chosen to 'disperse', the number of loci used, whether loci had high or low Polymorphic Information Content and the location to which the dispersers were moved. A review of the literature revealed that studies that have implemented first-generation migrant detection to date have used on average 10 microsatellite loci. Our results suggest at least 27 loci are required to accurately identify dispersers in the study system evaluated here. We suggest that future studies use the approach we describe to determine the appropriate number of markers needed to accurately identify dispersers in their study system; the unique nature of natural systems means that the number of markers required for each study system will vary. Future studies can use Simulated Disperser Analysis on pilot data to test marker panels for robustness to contemporary dispersal identification, providing a powerful tool in the efficient and accurate design of studies using genetic data to estimate dispersal.

Keywords: Ecological Genetics; GeneClass2; Migrant; Population Genetics; Power Analysis.

Figure 1. Probability of correctly identifying a simulated disperser by the number of loci used.

Genetic assignment tests were used to determine whether three common starlings (Sturnus vulgaris) from three genetically distinct collection localities would be identified as dispersers when their movement to a new location was simulated. Shown is the relationship between the probabilities (y-axis), from a Generalised Linear Mixed Model, of correctly identifying a simulated disperser plotted against an increase in the number of loci used. The top line indicates the response when using loci with the highest Polymorphic Information Content (PIC) first (blue line), and the bottom line indicates the response when using loci with the lowest PIC first (red line). The coloured shading corresponding to each line encompasses the upper and lower confidence intervals of the models.

Figure 2. Heatmap showing the assignment status of each simulated disperser.

Genetic assignment tests were used to determine whether three common starlings (Sturnus vulgaris) from three genetically distinct collection localities (A, Munglinup; B, Mallala; C, Orange) would be identified as dispersers when their movement to a new location was simulated. Each small coloured rectangle represents the results of a genetic assignment test; blue, individual was correctly assigned to its collection locality; red, individual was not recognised as a disperser or resident; orange, individual was identified as a disperser but assigned to the incorrect collection locality. The three panels represent the results for simulated disperser A, C and B. Within each panel the top heat map shows the results when first using loci with the highest Polymorphic Information Content (PIC) and the bottom heat map shows the results when first using loci with the lowest PIC. The grey arrows below each inset show the direction the loci were sorted, from high PIC to low PIC (top) or vice versa (bottom). Each inset is sorted on the y-axis by the location the individual was found and the movement treatment. Each row corresponds to a different treatment. The treatment order of each simulated migrant’s heat map row can be found in Table S4.

Figure 3. Data extracted during a literature review showing the relationship between the number of loci used and F_ST value.

A literature review was conducted to gather data on studies that used genetic assignment tests in GeneClass2.0 and their global F_ST. This graph shows the non-significant relationship (GLM, t₇₁ =  − 1.312) between the number of loci used and global F_ST of 72 datasets. The grey band shows the 95% confidence interval on the fitted values.

Figure 4. Number of papers that cited Piry et al. (2004) and contained the search term ‘microsatellite’ each year between 2005 and 2016.

Figure 5. Recommended workflow indicating the steps necessary for validating genetic assignment tests.