Analysis of SNP data, LD and population structure

The 133 soybean landraces had been genotyped with the 180 K AXIOM Soya SNP array previously [41]. According to MAF > 0.05 and missing data < 10%, we detected a total of 82,187 SNPs that were used for association mapping. LD was calculated using 44,838 SNPs covering the 20 chromosomes by PLINK V1.07. The pairwise LD (r²) among SNPs was estimated using E (r²) = 1/ (1 + 4N_ec) [59], where c represents the recombination rate of Morgan units and N_e represents the effective population size. The LD decay rate of the population was measured as the chromosomal distance when the average r² decreased to half its maximum value [23]. The STRUCTURE 2.3.4 software based on the Bayesian model was used to explore the population structure of the 133 soybean landraces based on 82,187 SNPs [41]. A total of 82,187 SNPs were employed to conduct principal component analysis (PCA) and construct a neighbor-joining phylogenetic tree using PLINK V1.07 and PHYLIP. The TASSEL V5.2.15 software was used to calculate the kinship matrix, which represents the similarity of the different pairs of SNPs between genotypes.