Adaptation Analysis
Nucleotide diversity (π) was calculated for the four groups using a 100 kb nonoverlapping sliding window by using VCFtools (Danecek et al., 2011). Additionally, the fixation index (FST) and nucleotide divergence (DXY) between each of the four groups were calculated by PIXY (Danecek et al., 2011) in 100 kb sliding windows. Ninety-five percent confidence intervals for mean FST and DXY values were obtained by bpnreg packages in R. PopLDdecay (C. Zhang, Dong, Xu, He, & Yang, 2019) software was applied to compute linkage disequilibrium (LD) decay among different groups and chromosomes separately. To make the following analyses easier to complete, we employed PLINK v.1.9 (Purcell et al., 2007) to filter SNPs with the following parameters: –indep-pairwise 50 10 0.2, 143,318 SNPs were kept and phased using Beagle v.3.3.2 (Browning & Browning, 2007). We ran Mantel tests to test for correlations between FST and geographic distance to assess isolation by distance (IBD) in the R package vegan (Oksanen et al., 2013). The QST were calculate using R package ‘QSTFSTComp’ (Gilbert & Whitlock, 2015).
To explore the effect of adaptation on 11 quantitative traits, we used the single-phenotype QST-FSTtest with the R package ‘QSTFSTComp’ (Gilbert & Whitlock, 2015). If QST > FST, the differentiation of traits is the major effect of divergent selection and shows local adaptation. We used the half-sib dam model and 10000 resampling steps for each QST-FST analysis. Taking into account the recent divergence between NE and EL lineages, we compared the levels of FST between CN and NW lineages to identify candidate loci under natural selection from 143,318 SNPs by BayeScan v.2.1 (Foll & Gaggiotti, 2008) software with default parameters, and PGDSpider (Lischer & Excoffier, 2012) was used to produce an input file for BayeScan. SNPs with a q value lower than 0.05 were considered potentially selected loci. Fisher’s exact test was used to perform significance in GO enrichment analysis on positively selected genes by the clusterProfiler package (Yu, Wang, Han, & He, 2012) in R.