Haplotype Computing Tools for Genomic Prediction and Estimation and Detection of Epistasis Effects in Holstein Cattle

Abstract

Genomic selection using genome-wide single nucleotide polymorphism (SNP) markers has become a widely accepted approach for genetic improvement in livestock and crop species. Haplotype analysis of genomic prediction and estimation uses information from genomic regions that could be chromosome regions or genes and is a step beyond the current genomic selection approach using single-SNP models. However, haplotype analysis is computationally challenging. This research developed a computing pipeline that made the analysis of haplotype genomic estimation and prediction a seamless data analysis process. This computing pipeline was applied to Holstein cattle and achieved limited success, unlike results from our parallel study on a human population where haplotypes using structural and functional genomic information improved the prediction accuracy for all seven human traits, and results from a collaborative study in swine where haplotypes of chromosome regions and genes improved the prediction accuracy for seven of the eight traits. We then conducted a large-scale genome-wide association study (GWAS) of epistasis effects in Holstein cows to investigate potential reasons for the limited success of haplotype analysis in Holstein cows. The results showed that complex intra-chromosome epistasis effects covering large and varying distances as well as inter-chromosome epistasis effects were the likely reasons for the limited success of haplotype analysis. Towards large-scale analyses of complex genetic mechanisms, a distributed, parallel computing version of the computing tool for haplotype analysis was developed to use distributed computing resources and to reduce the computing time for large-scale haplotype genomic prediction and estimation.