Abstract

In this report, we compare and contrast three previously published Bayesian methods for inferring haplotypes from genotype data in a population sample. We review the methods, emphasizing the differences between them in terms of both the models ("priors") they use and the computational strategies they employ. We introduce a new algorithm that combines the modeling strategy of one method with the computational strategies of another. In comparisons using real and simulated data, this new algorithm outperforms all three existing methods. The new algorithm is included in the software package PHASE, version 2.0, available online (http://www.stat.washington.edu/stephens/ software.html). Current high-throughput genotyping technologies, when applied to DNA from a diploid individual, are able to determine which two alleles are present at each locus but not the haplotype information (that is, which combinations of alleles are present on each of the two chromosomes). Knowledge of the haplotypes carried by sampled individuals would be helpful in many settings, including linkage-disequilibrium mapping and inference of population evolutionary history, essentially because genetic inheritance operates through the transmission of chromosomal segments. Experimental methods for haplotype determination exist, but they are currently timeconsuming and expensive. Statistical methods for inferring haplotypes are therefore of considerable interest. In some studies, data may be available on related individuals to assist in this endeavor, but in general such data may be either unavailable or only partially informative. We focus here on the problem of statistically inferring haplotypes from unphased genotype data for a sample of ("unrelated") individuals from a population. Several approaches to this problem have been proposed, notably Clark's algorithm