Submitted to: Journal of Animal Science Supplement
Publication Type: Abstract Only
Publication Acceptance Date: April 24, 1997
Publication Date: N/A
The objective was to develop a quantitative method of identifying potential errors in genetic marker data. Scoring errors (SE) cause problems in estimation of map distance, marker order and the effects and positions of quantitative trait loci. Several programs are available for identifying likely SE based on the presumption that marker data represents genotypes. They generally identify inconsistencies but often do not indicate which animal is responsible. An alternative approach is to view the marker data as phenotypes which arise from unobserved genotypes at the marker locus. The model incorporates a prior error probability. Consequently, errors have less effect than when genotypes are assumed to be observed without error. Posterior error probabilities (PEP) are calculated by summing the posterior genotype probabilities (PGP) of all genotypes that are inconsistent with the observed marker phenotype. The PGP for an individual (I) can be partitioned into prior allele probabilities for I's sire (S) and dam (D) conditional on the marker data related to I throughout S or D, respectively, and a likelihood of marker data on I or related to I through I's progeny. The priors and likelihood can be calculated recursively in terms of one another. PGP are computed by iteratively calculating priors from the top of the pedigree to the bottom and likelihoods from the bottom of the pedigree to the top.