Computational protocol: Flying between Sky Islands: The Effect of Naturally Fragmented Habitat on Butterfly Population Structure

Similar protocols

Protocol publication

[…] The spatial autocorrelation statistic r, which is closely related to Moran’s I, was estimated using the methodology from , implemented in GenAlex . Like Moran’s I, r also assumes values ranging from –1 to +1. Eight uniform distance classes of 100 km each was used for the MP dataset, five distance classes of 50 km each for the HO dataset, and six distance classes of 10 km each for the HO-ANA dataset. The significance of r in each distance class was established using 1000 permutations . The result of the permutation tests was analyzed using both one-tailed and two-tailed tests. The differences between the two datasets were estimated using three statistical tests, following the methodology described in Banks (2012) . [...] Since Bayesian clustering algorithms like STRUCTURE make a lot of assumptions (Hardy-Weinberg equilibrium, linkage equilibrium and regular sampling) while performing the clustering of genotypes , multivariate clustering approaches were adopted. A recently developed protocol, the spatial principal component analyses (sPCA) technique, was used . sPCA uses a synthetic variable derived from variation in allele frequencies and spatial information. Spatial information is usually provided in the form of a connection network, chosen based on the sampling protocol. Since sampling was clumped for MP and HO datasets, the inverse distances connection network was used; for HO-ANA, since the sampling was more evenly placed, a Gabriel graph connection network was used .Two permutation tests were used to test for positive and negative spatial autocorrelation: the Gtest (test for global structures, positive spatial autocorrelation), and Ltest (test for local structures, negative spatial autocorrelation). The PCA was performed in the package ade4 version 1.4 , and sPCA in the package adegenet version 1.3 , both implemented in the statistical language R v2.14 . [...] The genetic differentiation between the groups in each dataset was estimated using both band-based and allele-frequency based FST , . AFLPSurv was used to estimate allele frequencies, as described earlier, and the allele frequency values were used in the calculation of FST. Pairwise genetic distances (Nei’s D, after ) were also estimated. The significance of the FST values was evaluated using a permutation test, with 1000 permutations. The program Arlequin was used to perform the Analysis of Molecular Variance (AMOVA), in order to calculate the molecular variance found among groups, within groups, within populations and within individuals. [...] Assignment tests were carried out in AFLPOP . A relatively high minimum log-likelihood difference (MLD) of 1 was used, to reduce chances of misassignment . The value for ε was chosen as 0.001, as per the default settings. When individuals were allocated to a population other than the one where they were sampled from, the AFLPOP simulation option was used to assess the probability of incorrect assignment, using 1000 random specimens. The simulation gives a P value of allocation to the second population; when this P value is low (<0.01), the chance of correct allocation to the second population is high. In some cases, the software cannot place an individual into any of the populations with high confidence; in such cases, the individual is considered an immigrant from outside the sampled metapopulations. […]

Pipeline specifications