Computational protocol: Population genomics shed light on the demographic and adaptive histories of European invasion in the Pacific oyster, Crassostrea gigas

Similar protocols

Protocol publication

[…] For technical reasons related to the quality and quantity of DNA, AFLP genotyping was only possible on 12 of the 16 available samples. Detection of AFLP outlier loci was performed using a version of FDIST2 software (Beaumont and Nichols ) modified for dominant markers (DFDIST program; http://www.rubic.rdg.ac.uk/∼mab/stuff), Bayescan v2.0 (Foll and Gaggiotti ) and the method of Bonhomme et al. (). Frequencies of null alleles were computed using the approach of Zhivotovsky (). In DFDIST, a mean ‘neutral’ FST value was calculated after trimming 30% of the highest and lowest FST values (see Gagnaire et al. ). The number of demes was set at 100, and 50 000 loci were generated by coalescent simulations under the finite island model. The maximum frequency of the commonest allele allowed was set at 0.99. BayeScan was used with default parameters for the chain and model, with 50 000 iterations. Using the approach of Bonhomme et al. (), a matrix of Reynold's distances was computed in AFLP-SURV 1.0 (Vekemans ); a population sampled in Japan was used as the out-group and 50 000 iterations were made. The chi-square-approximated P-values were corrected for multiple testing according to the Benjamin–Benjamini-Hochberg method (as suggested in Bonhomme et al. ).Detection of microsatellite and SNP outlier loci was performed using a version of DFDIST modified by r. Vitalis to simulate codominant, bi-allelic data (see Ségurel et al. ); the method of Excoffier et al. (), implemented in Arlequin 3.5 (Excoffier and Lischer ); the method of Foll and Gaggiotti (), implemented in Bayescan v2.0; and the method of Bonhomme et al. (). For the last two methods, the parameters were the same as for the AFLPs. In R. Vitalis' modified version of DFDIST, 50 000 simulations were performed with 100 demes. The maximum frequency of the commonest allele allowed was set at 0.99. Because the finite island model has recently been shown to lead to a large fraction of false positives if populations are hierarchically subdivided, we used the modified version of FDIST implemented by Excoffier et al. () for codominant data that use a hierarchical island population model (as defined by Slatkin and Voelm ). Two groups of 100 demes were used (following the genetic structure results), that is, northern populations (Danish Wadden Sea, Isefjord, Kristenberg, Limfjord, Munkmarsch and Tjarno) and southern ones (Arcouest, Cap d'Agde, Dutch Wadden Sea, Grevelingen, Marennes, Normandie, Oosterschelde, Quiberon, Squiffiec and Japan). We used 50 000 coalescent simulations for the hierarchical model.For all methods and types of markers, outlier detection was performed on all populations and on southern populations and northern populations separately. The threshold for outlier detection with BayeScan was set at FDR = 0.05 for both AFLP and SNP. […]

Pipeline specifications

Software tools BayeScan, Arlequin
Application Population genetic analysis
Organisms Crassostrea gigas