Population Structure and Linkage Disequilibrium in Elite Soybean Breeding Germplasm from Minnesota

Characterizing population structure and linkage disequilibrium (LD) are essential steps in genomic breeding studies. The objectives of this study were to: (i) investigate existing population structure, (ii) calculate a SNP-based kinship matrix, and (iii) assess the extent and genome-wide distribution of LD in the Minnesota soybean breeding germplasm collection. A panel of 280 representative breeding germplasm were selected based on the sum of the weighted contributions in the University of Minnesota soybean breeding program and then genotyped by a genome-wide panel of 1536 SNP markers using Illumina next generation technology. Population structure was assessed using the program STRUCTURE and Principal Component Analysis (PCA) based on simple matching of SNP alleles in TASSEL (version 3.0). A matrix P, which is composed of the first 3 principal component vectors accounting for population structure, was generated as was a SNP-based kinship matrix calculated in TASSEL. LD as measured by (D’) among linked and neighboring markers was calculated in the program Haploview 4.2 and presented as a heat map. The results from both STRUCTURE and PCA indicated the availability of strongly differentiated subpopulations in the Minnesota germplasm panel. Three distinct subgroups were detected using STRUCTURE, which is in agreement with existing knowledge about the germplasm collection. The genome-wide distribution of LD also revealed that LD varied greatly throughout the genome, providing background information for genome-wide association studies. The ultimate goal of our project is to identify SNP-based marker-trait associations in the germplasm collection enabling soybean breeders to implement genomic selection to manipulate important traits as part of variety development. Thus, population structure, kinship and LD information from this study will be integrated into association genetics studies to identify valuable alleles and to train statistical models for predicting genomic estimate breeding values in Minnesota’s soybean breeding program.