Figure 1.

Histograms of the average error rates based on either individual phase calls (open bars) or the proportion of incorrectly inferred loci (shaded bars), for ACE (A), β₂AR (B), CFTR (C), and coalescence-simulation (D) data. For the ACE data, there are a total of 52 biallelic markers for 11 subjects (Rieder et al. 1999), and 100 independent runs for each algorithm were performed. For the β₂AR data, 15 haplotype pairs (each pair corresponding to one subject) were randomly drawn from a total of 10 distinct haplotypes according to their respective frequencies, as shown by Drysdale et al. (2000); this procedure was repeated to generate a total of 100 simulated data sets. For the CFTR data, the 100 data sets were generated by randomly pairing 56 of the 57 complete haplotypes of the 23 linked SNPs in a 1.8-Mb region near the CFTR gene provided by Kerem et al. (1989). The coalescence simulation was done using the Long Lab’s algorithm. A total of 100 replications were performed for a regional size of 10 U of 4Nc, each of which consisted of 20 pairs of unphased chromosomes with 20 linked SNP loci. The error bars are shown as ±1 SE, for the new version of PHASE, HAPLOTYPER, and PL-EM. An asterisk (*) indicates that the old version of PHASE was used for this data set, because its performance is better than that of the new version.