Use of SNP Marker Data to Analyse Population Structure in Crossbred Sheep

Abstract

Accurate prediction of breed composition in livestock species is important for a range of decision making in breeding programs. High density data on single nucleotide polymorphism (SNP) markers should provide more reliable predictions of breed compositions than data from microsatellite makers, which cannot achieve such high information content across the whole genome. Sheep populations in Australia consist of pure breeds, relatively recently developed synthetic breeds and various structured crosses. The main objective of this study was to assess the utility of SNP data to analyse population structure and breed composition in Australian sheep. First filial generation (F1) progeny from the Sheep CRC Information Nucleus (INF) were used in the analysis along with relevant breeds selected from the international Sheep HapMap data. After quality quntrol (QC) filtering, 46,700 SNPs genotyped on the Illumina Ovine 50k SNP chip were included in the study. Principal component analysis (PCA) was conducted with selected Sheep HapMap animals to determine the genetic structure of the pure breeds. Based on the PCA results pure breed individuals were selected as a reference population to obtain estimates of the breed composition of F1 offspring using STRUCTURE. Five crossbred groups were subjected to separate analyses with STRUCTURE using different subsets of SNPs and estimates were compared to the F1 pedigree expectation of proportion Merino (= 0.5). While animals from well-defined pure breeds formed tight clusters in the PCA, animals from synthetic breeds exhibited higher variation forming less well defined clusters. The F1 from the pure breeds were tightly clustered mid-way between the parent breeds while greater dispersion was observed in the F1 with a synthetic parental breed. When using all 46,700 markers in the analysis, the standard deviations for the estimated Merino proportions in the F1animals were 0.016, 0.026, 0.016 and 0.017 for, respectively, Merino X Border Leicester, Merino X Texel, Merino X Coopworth and Merino X Suffolk. With a higher number of SNPs the estimation of breed composition is highly accurate and has a low bias. The biases of estimation of breed proportion are higher with genetically more diverse synthetic breeds such as Coopworth and Suffolk.