384 Genetic and Genomic Analysis in Livestock with Increasing Datasets

Abstract

Abstract In animal breeding and genetics, statistical methods have been used for several decades to identify animals with the best genetic potential. One of the foundations for computing accurate estimated breeding values (EBV) is the amount of data that is used in the evaluation system — as the more data points one animal has, the more accurate its EBV is going to be. However, the animal breeding and genetics field periodically faces a big data paradox, where efficient methods have to be developed to handle the amount of data collected over time, given the computing capacity becomes the limiting factor. For instance, running genetic evaluations based on phenotypes and pedigree for a million animals was impossible in 1970. Methods and algorithms evolved to a point where using data for millions of animals was not a problem, until genomic information became available. After the development of single nucleotide polymorphism (SNP) chips for livestock in 2008, genomic information started being used in addition to phenotypes and pedigree to further improve accuracy of EBV. However, each animal is genotyped for around 50,000 SNP, which makes this data dense and difficult to work with. Over 790,000 Angus and 3.4 million Holstein cattle have been genotyped in the US as of April 2020. As the amount of new data considerably increases every week, most of the genomic evaluations are done on a weekly basis. Given that data have to be processed, the computation of EBV cannot take more than four days, which can be challenging depending on the model. In this talk we will discuss the challenges and solutions for successful genomic evaluations in large livestock populations. Finally, perspectives on the use of whole-genome sequence data and high-throughput phenotypes in genomic analysis will be summarized.