Population Genetic Structure. Two Bayesian, model-based approaches were taken to provide basic information regarding raccoon genetic structure, specifically in the context of avoiding undetected population structure that might obscure the results, and to examine potential differences in rabies infection if more than one population was found (i.e. potential local adaptation). The first method used Structure (Xxxxxxxxx et al. 2000) to identify the most probable number of populations in the study area. Structure can be used to analyze multi-locus genotypic data alone or in conjunction with geographic information, albeit indirectly. In the latter case, the user assigns individuals to “K” populations based on their geographic sampling locations and tests to see if the population assignments reflect geographic structure. Since Structure cannot directly include latitude/longitude coordinates, the analysis was based solely on the genetic data. Simulations were performed in triplicate with 2,000,000 iterations (100,000 dememorization steps) for 1-5 possible populations
Population Genetic Structure. Each of the replicate Structure runs resulted in similar output values, indicating overall parameter stability. Based on the posterior probability values calculated from the data, the highest support was found for 2 raccoon populations in northeastern Ohio (Pr(K=2) = 0.82; Table 2.3). Table 2.3 Number of populations (K) estimated using Structure. The posterior probability of the number of populations is given by Pr(K), which is determined using the natural log of the probability of the data (X) given the number of populations (K), ln Pr(X|K). 1 -4198.8 0.18 2 -4197.3 0.82 3 -4205.3 0.0002 4 -4229.0 1.39 x 10-14 5 -4285.9 4.95 x 10-39 One hundred raccoons were assigned to population 1 whereas 82 raccoons were assigned to population 2 (Figure 2.2). Three individuals were equivocal with respect to their population assignment. The two potential populations overlapped geographically in more than 30% of the sampling area. Alternatively, the Geneland analysis identified a single population in the study area (Figure 2.3) based on the average posterior density of the 3 simulations (0.6413). Geneland indicated that the next most likely scenario was the existence of 2 populations, however, the average posterior density of the 3 simulations was much lower (0.1807) for this situation. In general, average posterior density values declined with increasing values of K (Figure 2.3). Figure 2.2 Allocation of raccoons into 2 populations, as suggested by the Structure analysis. Population 1 (blue) contained 100 raccoons, and population 2 (red) contained
Population Genetic Structure. An accurate representation of raccoon population genetic structure is necessary to discern between alternative hypotheses; therefore, Bayesian model-based clustering algorithms implemented in two programs, Structure (Xxxxxxxxx et al. 2000) and Geneland (Xxxxxxx et al. 2005), were used to determine the most probable number of raccoon populations in the eastern US, in the areas surrounding the hypothesized suture-zones. One potential disadvantage of the Structure program is that in certain instances, linkage disequilibrium, deviations from Xxxxx-Xxxxxxxx equilibrium, null alleles, isolation by distance, and differences in sample size between clusters can all confound the analysis and lead to a potential overestimation of populations or genetic clusters (Xxxxxxxxx et al. 2007; Hubisz et al. 2009); therefore, use of a second method like the spatially-explicit Geneland is recommended. Although Structure can incorporate geographic information indirectly, it is a crude measure whereby the user assigns individuals to K populations based on their geographic location, and the resulting output is compared with what one might expect if there was indeed geographic structure. Therefore, Structure is generally regarded as a nonspatial method. As a result, simulations solely on the genetic data were performed in triplicate with 5,000,000 iterations (500,000 dememorization steps) for a possibility of 1 to 10 populations (K) under a model of admixture. The average natural log of the probability of the data for each possible number of populations was then used to estimate the posterior probability of the most likely number of populations based on Bayes’ Rule, as described in the Structure user guide (Xxxxxxxxx et al. 2007). Geneland, on the other hand, can simultaneously incorporate spatially explicit data (i.e. latitude/longitude coordinates) and genetic data during computations. Therefore, Geneland runs were based on the microsatellite data and an associated spatial location (latitude/longitude) for each sample. As above, simulations were performed in triplicate with 5,000,000 iterations (500,000 dememorization steps) for a possibility of 1 to 10 populations (K). The most probable number of populations was then chosen based on the posterior density averaged over the 3 Geneland runs. After determining the most probable number of populations, an Analysis of Molecular Variance (AMOVA) was used to examine levels of genetic differentiation between the populations and to confirm...
