Characterizing allelic associations from unphased diploid data by graphical modeling

A method for estimating a graphical model to describe allelic associations between genetic loci is extended to use diploid genotypes rather than haploid data. It also provides haplotype frequency estimates, estimates of phase for sampled individuals, allows for missing or partial information, genotyping errors, and arbitrary penetrance functions. A data set of 688 unrelated individuals genotyped on 25 genetic loci is used to illustrate haplotype frequency estimation. The frequencies obtained are shown to be similar to those obtained by the PHASE program. We also illustrate how putative loci for traits can be included in the analysis in order to detect allele-phenotype associations. Haplotype reconstruction is illustrated on a standard set of 40 male X-chromosome haplotypes, randomly paired to simulate diploid genotype data. The novel method is shown to be comparable to most existing ones, though the PHASE method does consistently better on this problem. The graphical model approach is shown to have some advantages in terms of tractability and can be used to select informative subsets of loci and to map loci influencing phenotypes.