Effect of genotype imputation on integrated model for genomic selection

Abstract

Genomic selection is a very recent area of study in case of molecular breeding of livestock or crop species. There are various statistical models available for genomic selection. The performances of these models depend on several factors like sampling population, genetic architecture of target species, statistical models as well as missing genotypes. Missing genotype is very common problem in high throughput sequencing data. These missing genotypes are necessary to be imputed in order to implement the genomic selection models. Different statistical models of genomic selection behave differently in imputed data. So, it is highly imperative to evaluate the performances of statistical models under different levels of imputations to know the behavior of the models. In this article, performance of three statistical models viz. Sparse Additive Models (SpAM), Hilbert-Schmidt Independence Criterion Least Absolute Shrinkage and Selection Operator (HSIC LASSO) and Integrated Model for genomic selection are compared after incorporating the various degree of imputation (0, 2, 5 and 10%) in the real data. Results indicate that integrated model is found to be more robust against the level of imputation of the genotypic data.