Bayesian Networks in Equine Genetic Counseling

Abstract

Evaluation of the individualized risks of occurrence of genetic disorders in horses can be challenging and necessitates adequate methods. One of these methods is the Bayes network (BN). The objective of this study is to create a BN and test its utility in evaluating risk occurrence of type 1 polysaccharide storage myopathy (PSSM1) as an example. Current knowledge on the hereditary mode of transmission of PSSM1 and on the effects of breed, diet, exercise regimen, and age on the disease was retrieved from the scientific literature. Uncertainty on this background information was converted into prior beta distributions. Utility values of different potential mating were based on recommendations toward preventing the production of affected animals and reducing the frequency of asymptomatic carriers. The network suggested, for example, it is impossible to observe parents free of PSSM1 clinical signs if their offspring present PSSM1 clinical signs. Expected utilities for the alternatives “breed” or “not to breed”, two individuals free of PSSM1 clinical signs, are equivalent if an arbitrary score of 50 is given to each copy of the deleterious allele. It is higher for the alternative “breed” if a score of 100 is given to a mating leading to a heterozygous offspring. Results of the sensitivity analyses agree with intuitive expectations about the usefulness of genetic test in determining the genetic status of an individual. Limited current knowledge about the risk factors of PSSM1 may affect the results but the network may be easily updated with incoming new information.