Deliverable D5.4 Methodologies to account for crossbred and genomic data in selection for feed efficiency

Objectives

The general objective of this work was to develop methodologiesto account for crossbred and genomic data for a sustainable selection for feed efficiency. These activities were essentially developed in task 5.4 of the Feed-a-Gene project. Specifically, we have worked on three main topics of research: 

1. Assess the contribution of the additive and dominance genetic effects to the phenotype expression of several traits, especially those related to feed efficiency, within crossbred and purebred animals, and the genetic correlation to the corresponding traits between the two populations, 
2. Identify optimal genomic prediction models, both in terms of efficiency in calculations and accuracy of estimates, that enable unbiased and accurate estimation of the genetic parameters and genomic breeding values that are required for implementation in practice, 
3. Combine all available and newly generated knowledge in terms of genetic variances and economic values to propose a new index to improve feed efficiency.

Rationale

Selection to improve feed efficiency in monogastrics has been practiced for several decades, despite the practical difficulties and costs related to routinely measuring feed intake and efficiency on a large scale. These selection strategies so far have mostly been based on measured feed intake and efficiency on purebred animals, because selection takes place within purebred lines. The breeding goal, however, is to improve feed efficiency in the crossbred production animals. Strategies to select purebred selection candidates based on performance measured on crossbred offspring or relatives have been proposed previously, but linking crossbred performance back to the purebred animals was often challenging, while the links were relatively weak. Genomic selection is the state-of-the-art selection approach that is used in modern breeding programs. The previously described limitations can be overcome by using genomic selection of purebred animals, based on a crossbred training population. This requires genomic prediction models that can compute genomic breeding values for purebred animals for crossbred performance. In such models, it is important to appropriately model the crossbred and purebred animals relative to each other. From the research undertaken in Task 5.4 of Feed-a-Gene, as well as from recently published research, a straightforward genomic prediction model that models at least one so-called metafounder for each of the parental lines, appears to be the most appropriate choice. Such a model provides genomic breeding values with accuracies similar to those obtained with more sophisticated models and similar or less bias than other models, while being one of the most computationally efficient models. Using genomic information in the estimation of breeding values provides the opportunity to explicitly model dominance deviations, next to the traditional additive breeding values. It was expected that explicitly modelling dominance deviations is especially relevant in crossbreeding, where the crossbred performance partly relies on heterosis, which is mainly caused by dominance effects. Within Task 5.4 of Feed-a-Gene, an empirical study was conducted that showed that 6-12% of the phenotypic variance of feed efficiency related traits in pigs is caused by dominance effects, while 18-30% is caused by additive effects. Based on these results, as well as recent studies that investigated the impact of explicitly modelling dominance deviations in genomic prediction models, it is concluded that the potential benefits from accounting for dominance is likely to be relatively small for improving feed efficiency and growth.

In WP5, several new traits and characteristics have been proposed and investigated, that have some relationship with feed intake and/or efficiency, and that may be easier to record than feed intake or efficiency itself. For each of those traits, the question is whether breeding programs in practice should consider measuring those and selecting on them. Therefore, the ultimate approach, as defined in Task 5.4, was to evaluate new selection indices, including those new traits. We showed that selection response for crossbred feed efficiency can be increased by: including an economic weight for the crossbred rather than the purebred trait, adding crossbred information for traits not measured directly on purebred selection candidates, and additionally adding indicator traits, of which especially digestibility, feeding behaviour, and biomarkers are beneficial. Including genomic prediction is also recommended, however this would require investment to maintain a reference population of crossbred pigs.