What if choosing the best pigs to breed, both phenotype and behavioral traits, was as simple as running them over a scale with a video camera?
While high-throughput pig phenotyping remains labor intensive and costly, a recent study set out to determine valuable, heritable traits while using less people and the addition of one camera during standard weighing routines.
The study focused on automated pose estimation, using the automated video-analytic system, DeepLabCut. The body dimension and activity trait data collected, along with pedigree information, expands on the standard routine of pig weighing and aims to estimate genetic parameters for breeding programs, specifically phenotype and behavioral traits, on an individual level.
After analyzing 7,428 recordings of 1,556 finishing pigs, body metrics were found to be highly heritable (61% to 74%), while activity traits were low to moderately heritable (22% to 35%) and showed low genetic correlations with production traits and physical abnormalities, the study explains.
Pigs’ body length, width and height are moderately to highly heritable (20% to 60%), and behavioral traits have also been shown to be heritable. However, heritability estimates differ considerably, likely due to manual scoring by observer bias, the study notes.
The system, DeepLabCut, aims to create reliable and transferable tracking models to quantify behavioral traits with limited observer influence.
The Study
During the study, measurements were taken individually of the body length, hip and shoulder width and surface area, using an eight-body part configuration, as well as activity scores of pigs during weighing. Each pig was also recorded and scored manually, noting physical abnormalities, including ear swellings or hemotomas, the presence and size of umbilical hernia, ear biting wounds and tail biting wounds.
Activity traits were concluded through movement of the pigs from entrance to departure on the weighing scale. Considering the speed and acceleration of the pig, as well as it’s “searching behavior,” which cause the pig to deviate from traveling in a straight line, helped determine the behavioral tendencies of the pigs. For example, fast moving pigs with strong “searching behavior” attributes, might be considered less docile animals, compared to a pig that is slower moving and is calm on the scale from beginning to end of the observation.
Automated body dimensions and activity traits were validated by comparing values with manual activity scores recorded by manual recordings and observers, respectively.
The Results
The DeepLabCut estimates of body dimensions were validated through manually collected body dimensions, while activity scores had moderate to high correlations with manual scores.
Results accurately displayed both pigs’ body dimensions and activity traits and help determine the heritability of these.
Could the ability to automatically determine such traits be the future for pig breeders and help advance the pork industry?
