960
Use of the canonical discriminant analysis for selecting a panel of informative markers in 21 Italian sheep breeds
SNP markers can be useful to assign individuals to their breed. In the present research, a high density SNP chip was exploited to select a panel of markers useful to verify the origin of 462 individuals belonging to 21 Italian sheep breeds. Animals were genotyped using the Illumina’s OvineSNP50 BeadChip and were divided into two groups: the training population (TP) of 420 sheep, 20 for each breed, and the validation population (VP) of 42 animals, 2 for each breed. SNPs with MAF<0.20 were discarded and, after data editing, 40856 markers were used, including 4937 monomorphic SNPs for at least one breed. The canonical discriminant analysis (CDA) was exploited to discriminate among breeds. CDA was first applied at chromosome level and markers with canonical coefficients higher than a fixed threshold were retained. The threshold was obtained through a recursive procedure: the CDA was repeatedly applied by increasing, at each run, the value at which a canonical coefficient is discarded. The procedure stopped when the remaining markers were a pool of linearly independent variables. A genome-wide CDA was then developed with only the selected SNPs and the effective distance among groups was tested by using the Mahalanobis’ distance and the corresponding Hotelling’s T-square test. The discriminant analysis (DA) technique was then used to assign the VP to the breed of origin. Finally, the minimum number of significantly discriminant markers was obtained. With a canonical coefficient threshold value of 0.31,, 155 linearly independent highly discriminant SNPs were retained . These selected markers provided at Hotelling’s T-square test significant separation among all breeds (P<0.0000) and the DA correctly assigned 40 out of 42 VP animals . Among the 155 markers, 46 were monomorphic for at least one breed. The selected markers could be used to develop an assay to routinely track monobreed products. Finally the minimum number of markers able to significantly discriminate all breeds was 48. However, by using this small panel of SNPs, the DA was able to correctly assign only 30 out of 42 VP animals.
Keywords: assignment test, SNP selection