45 Quantifying gait quality changes in fragile foal syndrome carriers using artificial intelligence

Abstract

The domestic horse has remarkable locomotor polymorphisms valuable to study of genetic musculoskeletal diseases and their effect on gait quality. Here we investigate a polymorphism in the PLOD1 gene, responsible for Fragile Foal Syndrome (FFS), as it impacts the organization of collagen fibers. We hypothesize that the changes in the biomechanical properties of tissues will result in change of locomotor phenotypes. We first addressed the need for a gait assessment tool that can accurately quantify various locomotor parameters. We created a pipeline built around the software package DeepLabCut (DLC), a user-friendly graphical user interface (GUI) written in Python, and a custom gait parameter calculator written in MatLab. DLC uses a pre-trained deep neural network model to apply 22 anatomical landmarks on jogging horses observed in video recordings. A custom GUI allows the user to perform error checks of the labels before the pipeline processes the raw marker coordinate files produced by DLC into geometric gait parameters. We then genotyped 138 horses in training for jumping competition by extracting DNA from the bulbs of pulled tail hair and running a PCR-RFLP test. We confirmed these genotypes via sanger sequencing. Of the 138 horses, 58 were purebred Thoroughbreds, 72 were warmbloods, and the remaining 8 were made up of stock horses. The carrier rate was n = 4 in Thoroughbreds, and n = 3in the warmbloods including an Irish Sport Horse, Sillia Argentina, and Holsteiner. We observed a FFS carrier frequency markedly higher in the Thoroughbred population than those reported in previous studies (6.9%, P = 0.015, Χ2 test). A Quadratic Discriminant Analysis then produced standardized coefficient scores for each of the 15 gait parameters to determine the best predictors of genotype within the model. The change in joint extension of both the hind and forelimb were the strongest predictors of genotype and indicated that FFS carriers had greater extension and swing of all 4 limbs. As a result, FFS carriers likely have an altered hoof flight path throughout the stride cycle. Future work will increase the sample size to capture more FFS carriers and will more precisely define the FFS locomotor phenotype. We will also apply this new phenotyping method to investigate changes in gait associated with other genetic variants for connective tissue disorders like HERDA and JEB. Improved quantification of gait parameters, and our understanding of the physiology of this and other known connective tissue disorders in the horse, will aid in future breeding and selection decisions to reduce injury risk and improve performance.