Epsilon Archive for Student Projects

Abstract

Information from thousands of markers distributed across the genome can be used for a new selection method in animal breeding and genetics. This method which is called genomic selection estimated the genomic breeding value based on the estimation of marker effects covering the whole genome. For a successful application of genomic selection, accuracy of the prediction is an important factor that should be considered. Because it is important for genetic progress. Quantitative trait nucleotides (QTN) are polymorphisms that give useful information about gene function and QTL architecture. So prediction its effects and estimation its accuracy enhances rates of genetic gain. In this study, we investigated the accuracy of QTN prediction by neighboring markers, using the simulated data. Our dataset consisted of 1040 markers which were assigned to one chromosome of 500 genotyped animals. Method G-BLUP was used to estimate marker effects, and the accuracy of QTN prediction was estimated by using cross validation. As the accuracy can be affected by different number of surrounding SNPs, it was predicted at various number of surrounding markers ranging from 10 to 100 markers. In general, the accuracy of QTN prediction increased by increasing the number of flanking SNPs from 10 to 60 SNPs. Further increase in number of SNPs resulted in a very small increase in accuracy in case of heritability 1 and 0.8 and a very small decrease in case of heritability 0.5. We also investigated the effect of other factors on accuracy such as Minor Allele Frequency cutoff threshold, heritability and number of phenotypes in the training set. We analyzed four data sets; data set with no selection of markers, data sets with different cutoff thresholds for MAF (0.02, 0.05 and 0.1) in order to get the effect of MAF on accuracy. We observed the minimum SNP MAF of 0.02 is more appropriate for genomic selection studies. After filtering the data with the cutoff threshold of 0.02 for MAF, QTN could be predicted with 100 flanking SNPs, with a maximum accuracy of 0.777. This is the maximum accuracy in the absence of any environmental effects. We also observed that there is a relationship between the accuracy of QTN prediction and the heritability of the phenotype. The accuracy of QTN prediction dropped when the heritability of phenotype decreased. In general, when we estimated the accuracy by 100 surrounding SNPs and heritability decreased from 1 to 0.8 and from 0.8 to 0.5, the decrease in accuracy was 4.6 and 11%, respectively. In another analysis, when 50% of animals were masked, it means that the number of phenotypes decreased in training set, the accuracies were lower in comparison to 20% masking. When 50% of animals were masked, with 100 surrounding SNPs, the reduction of 6 and 9.25% was observed, when heritability of phenotype was 1 and 0.5, respectively.