[Evaluation of computational methods for HLA three loci haplotype by compare with family-based data]

We evaluated the accuracy and efficiency of computational inference methods for haplotype on estimate HLA-A-B-C haplotype frequencies by compared with the haplotypes manually defined in a family-base dataset.

558 individuals with pedigree information were selected, and their haplotyps were compared with the data obtained by the following three method: the Expectation-Maximization (EM) and Excoffier-Laval-Balding (ELB)algorithms using the AELEQUIN software, and the SAS/Genetics PROC HAPLOTYPE method.

After performing the SAS/Genetics method, and the Expectation-Maximization (EM) and Excoffier-Laval-Balding (ELB) algorithms using the AELEQUIN software, 248, 247, and 238 different haplotypes were obtained respectively. The accuracy rates of these three methods were 88.5%, 89.1%, and 90.3% respectively. There are no significant different in the accuracy and estimated haplotype frequency comparisons among any two of these computational inference methods.

High accuracy haplotype frequency estimate rates could be obtained by these three computational inference methods, and there are no significant difference in the comparison of haplotypes estimated by SAS/Genetics, the EM and ELB algorithms using the AELEQUIN software. However, ELB algorithm shows better performance than EM algorithm and SAS/Genetics PROC HAPLOTYPE method for haplotype frequencies estimation in general.