Optimizing genomic breeding programs

Availability of genomic data have revolutionized animal breeding by the implementation of genomic selection but there are still many questions unanswered, e.g., how contemporary breeding programs can be reshaped for optimization. This projects aims to develop methodology for improving animal breeding programs irrespective of the species. The to be developed methods will be most relevant for quantitative traits. The project comprises two main topics: 1) Utilizing mendelian sampling variance (MSV): The most widely used criterion for selection in contemporary breeding programs is the estimated breeding value (EBV). A BV is a measure of the genetic potential that an animal passes on to progeny on average. Basing selection on the EBV aims to increase the mean performance of the offspring generation. However, breeding programs are not (necessarily) interested in the mean of all animals of the next generation. They are rather interested in the mean BV of animals that will be selected as parents in the next generation, i.e., a subset of the offspring generation. Thus, the objective of this topic is to develop a criterion that is designed to selected animals with regards to their ability to produce good parents in the next generation. For that, the MSV of a mating will be used. 2) Optimizing parameters of breeding programs: Countless parameters are used in breeding programs, e.g. selection intensity, selection accuracy etc. It is unknown whether the currently used values for parameters are the best ones to meet predefined objectives. Thus, the objective of this second topic is to develop a method to find these best values through an intelligent search of the high-dimensional parameter space and evaluating parameter combinations by means of simulation. The intelligent search will be implemented by using Bayesian Optimization. Bayesian optimization is chosen because of its high flexibility, as it for instance does not require assumptions about the distribution of the parameter landscape.