Computational protocol: Effects of Large-Scale Releases on the Genetic Structure of Red Sea Bream (Pagrus major, Temminck et Schlegel) Populations in Japan

Similar protocols

Protocol publication

[…] Levels of genetic variation were characterized by counting observed allele (A), allelic richness (A r) and gene diversity within samples (H S) and the average for all samples (H T), based on Nei and Chesser [], using FSTAT v. software package []. Deviations from Hardy–Weinberg (HW) equilibrium, including significantly higher or lower F IS [] estimates than expected by chance, were investigated by Fisher’s exact probability test in GENEPOP v. 4.0 []. We adopted the false discovery rate (FDR) approach [] when interpreting the significance of test results. Linkage disequilibrium (LD) among all pairs of loci in all samples was tested by a Fisher’s exact test with 10,000 demorizations, 100 batches, and 1000 iterations per batch in GENEPOP v. 4.0 []. The presence of null alleles or technical artifacts was investigated with MICROCHECKER v. 2.2.1 []. Statistical evidence for selection was tested for by the outlier tests implemented in BAYESCAN [].Genetic differentiation among samples were quantified by Wright’s F ST, using Weir and Cockerham’s [] estimator θ within all samples and also within pairs of sample localities. The statistical significance of the analysis was estimated by exact tests in GENEPOP v. 4.0 [], with 10,000 dememorizations and batches, using 10,000 iterations per batch. The p values were calculated for each locus separately and summed over loci by Fisher’s summation procedure following Ryman and Jorde []. In this case, we adopted the Benjamini and Yukutieli's [] FDR approach, which is applicable also to non-independent tests, when interpreting the significance of p values. [...] Spatial genetic differentiation patterns were examined by a Principal Component Analysis (PCA) based upon a covariance matrix of allele frequencies, and visualized using PCAGEN v. 1.2.1 []. The Bayesian clustering method implemented in the software STRUCTURE v. 2.3.3. [] was performed to characterize spatial patterns of genetic clusters (K) and to infer admixture proportions between wild and hatchery-reared fish in the data set without a priori information of population partition. We assumed an admixture model and correlated allele frequencies []. Each run consisted of a burn-in of 50,000 MCMC steps, followed by 200,000 steps, for values of K between 1 and 7, and the calculation was done five times for each K. The most likely number of clusters, K, was estimated as the value which maximized the averaged log-likelihood, log Pr(X|K) and the ad hoc statistic ΔK []. Once K was determined, individuals were assigned to the respective clusters and plotted with DISTRUCT v. 1.1 [].Geographic patterns and the scale at which genetic structuring occurs were further investigated by testing putative correlations between genetic and geographic distances []. Pairwise F ST estimates were linearized, as F ST ⁄ (1—F ST), and regressed against the natural logarithm of the shortest maritime distance connecting each sample pair (). Isolation-by-distance effects were tested by a Mantel test performed in IBDWS v. 3.23 ([], As sample KHR comprised only hatchery-released specimens, it was excluded in this analysis and in the subsequent landscape genetic analysis performed in BARRIER v. 2.2 []; this software was used to identify barriers to gene flow among locations. As input for the program, we used sample geographical coordinates and pairwise F ST estimates along all pairs of localities. The analysis was performed including all loci to infer the rank of importance of the barriers. Statistical support for each barrier was evaluated by the number of loci that supported it, and by 1000-bootstrap analysis of the multilocus Weir and Cockerham’s [] F ST matrix generated using the DIVERSITY R package [].The amount of genetic variation explained by alternative sample grouping was tested with AMOVA using ARLEQUIN v. 3.5 []. As all samples included adults of different length ranges; we did not consider temporal effects in relation to the year they were collected, i.e., 2007 or 2008. We specifically tested the null hypotheses of panmixia, as well as of structuring by geographical regions (Japan Sea, Pacific Ocean and Seto Inland Sea) and by the PCA analysis results (KHR, KKB, KSB, WAK and the rest of the samples).We estimated the effective population sizes (N e) under a sample-size bias correction [] using the linkage disequilibrium method implemented in LDNE v. 1.31 []. Harmonic mean N e and jackknife-adjusted 95% confidence intervals for each sample were estimated using Pcrit = 0.02, i.e., excluding those alleles with frequencies lower than 0.02. Waples and Do [] recognized the limitations of the method in obtaining precise estimates when populations are large; still, they found the approach useful to discriminate between small and large populations. […]

Pipeline specifications

Software tools Genepop, BayeScan, DISTRUCT, IBDWS, Arlequin
Databases KKB
Application Population genetic analysis
Organisms Abramis brama, Pagrus major