Mapping genomic regions affecting milk traits in Sarda sheep by using the OvineSNP50 Beadchip and principal components to perform combined linkage and linkage disequilibrium analysis

BACKGROUND: The detection of regions that affect quantitative traits (QTL), to implement selection assisted by molecular information, remains of particular interest in dairy sheep for which genetic gain is constrained by the high costs of large-scale phenotype and pedigree recording. QTL detection based on the combination of linkage disequilibrium and linkage analysis (LDLA) is the most suitable approach in family-structured populations. The main issue in performing LDLA mapping is the handling of the identity-by-descent (IBD) probability matrix. Here, we propose the use of principal component analysis (PCA) to perform LDLA mapping for milk traits in Sarda dairy sheep. METHODS: A resource population of 3731 ewes belonging to 161 sire families and genotyped with the OvineSNP50 Beadchip was used to map genomic regions that affect five milk traits. The paternally and maternally inherited gametes of genotyped individuals were reconstructed and IBD probabilities between them were defined both at each SNP position and at the genome level. A QTL detection model fitting fixed effects of principal components that summarize IBD probabilities was tested at each SNP position. Genome-wide (GW) significance thresholds were determined by within-trait permutations. RESULTS: PCA resulted in substantial dimensionality reduction, in fact 137 and 32 (on average) principal components were able to capture 99% of the IBD variation at the locus and genome levels, respectively. Overall, 2563 positions exceeded the 0.05 GW significance threshold for at least one trait, which clustered into 75 QTL regions most of which affected more than one trait. The strongest signal was obtained for protein content on Ovis aries (OAR) chromosome 6 and overlapped with the region that harbours the casein gene cluster. Additional interesting positions were identified on OAR4 for fat content and on OAR11 for the three yield traits. CONCLUSIONS: PCA is a good strategy to summarize IBD probabilities. A large number of regions associated to milk traits were identified. The outputs provided by the proposed method are useful for the selection of candidate genes, which need to be further investigated to identify causative mutations or markers in strong LD with them for application in selection programs assisted by molecular information.