Use of different statistical approaches to study selection signatures in sheep breeds farmed in Italy

Natural and artificial selection affect genome structure causing genetic variation between breeds. Dense marker maps of thousand SNP disseminated across the whole genome allow for the investigation of chromosomal regions that differ between breeds. Several statistical approaches have been proposed to study selection signatures in livestock species. In this work, four approaches were used to study selection signatures in a sample of 496 sheep belonging to 20 Italian breeds, different for geographical origin and production aptitude. The four approaches were: I) Fst Outlier Detection (FOD), implemented in the LOSITAN software. II) comparison of Breed LS means of the sum of differences in SNP allele frequencies along sliding windows (SNP_DIFF). III) Correspondence analysis (CA). VI) Canonical Discriminant Analysis (CDA). Animal were genotyped with the Illumina OvineSNP50 BeadChip. The first six chromosomes were considered. After data editing, a total of 20,194 SNP were retained for the analysis. The different approaches were able to identify the same regions expressing variation between breeds. On OAR6, for example, all methods highlighted a region located between 35 and 41 Mb, where BMPR1b and ABCG2 loci map. Moreover, SNP able to differentiate between breeds were also detected at 76, 96 and 107 Mb, near to KIT, IL8 and SCD5 loci, respectively. All methods were able to discriminate breeds and, in general, a geographical pattern of variation has been detected. However each approach may supply different kind of information. FOD detected a relatively low number of markers in divergent selection but it was able to identify loci under balanced selection. CA and CDA decomposed the total variability of SNP markers among breeds in different and uncorrelated variables that could be useful for the identification of genes influencing complex traits.