Relationship inference from trios of individuals, in the presence of typing error

Misspecification of relationships and of genotype data can cause problems i n linkage analyses based on genome-scan data. Previous reports have focused on pairwise relationships and a simple error model. This article considers the increased information available from the joint analysis of trios of in dividuals, integrating this analysis with an error model that allows for th e most common genotyping errors. Given observed marker phenotypes in a geno me scan, computational methods are outlined both for likelihoods of relatio nships and for the posterior probabilities of underlying genotypes. The met hods are applied to examples from two real data sets: one has been previous ly well analyzed, and, hence, Mendelian inconsistencies have been removed; the other typifies the pedigree and genotype errors encountered in the init ial analyses of a study. It is demonstrated that the coupling of relationsh ip inference and error detection is quite effective, that the error model i s computationally practical, and that data on a third relative can often cl arify relationships.