#17-023

Complete

Category

Date Full Report Received

04/30/2020

Date Abstract Report Received

04/30/2020

Investigation

Institution:
Primary Investigator:
Co-Investigators: Janice Siegford, Ronald Bates, Catherine Ernst

This project focused on measuring and modeling behavior to enable selection for decreased aggression in group housed pigs. In order to select and breed less aggressive pigs, it is necessary to: 1) measure aggression in thousands of pigs, 2) incorporate those measurements into models for estimation of breeding values. For the automatic detection of aggression (objective 2) we first compared several programs to track animals, using existing video recordings of group housed pigs, to assess if the track of an individual animal was informative of its behavior. All tested programs failed to track animals for more than a few seconds or minutes. Thus, we moved to develop programs to directly detect aggression. In collaboration with the Catholic University of Leuven (Belgium) we develop an algorithm that can detect fights in group-housed pigs. The algorithm was applied to over 1600 3-second long video segments of group housed pigs, manually labeled as “aggressive interaction” (when pigs were fighting in the segment) or “non-aggressive interaction” (when pigs were not fighting, and the algorithm correctly classified 97.5% of the video segments. But measuring a trait (in this case detecting and quantifying aggression) is only the first step in implementing selection for such trait. Moreover, it is necessary to incorporate the recorded data into a genetic evaluation model. In Objective 1 of this project we improved existing models for selecting animals that are better suited for production systems that use group-housing of animals. We used a model that estimates the effect that each animal has on the phenotypes of other animals and we expanded it to incorporate behavioral measures. Our proposed model has several appealing features: 1) It allows modeling other traits including performance (e.g: growth rate, feed efficiency, etc) and welfare (e.g: skin lesions) as a function of the aggression between pairs of animals. 2) It can be used to select animals that perform well in groups (called direct genetic effect) and that do not affect the performance of the rest of the group members (called social genetic effect). In this way, we can, for instance, select an animal that not only fights less with its pen mates, but that also, when it fights, creates less skin lesions and that does not affects negatively the growth of other pigs in the group. We showed that by incorporating the behavioral measures in our model we could recover up to 50% more genetic variance, which in practice will result in more effectively selecting for pigs better adapted to group-house production systems.