Accuracy of programs for the determination of HLA alleles from NGS data

33 34 The human leukocyte antigen (HLA) genes code for proteins that play a central role in the 35 function of the immune system by presenting peptide antigens to T cells. As HLA genes show 36 extremely high genetic polymorphism, HLA typing on the allele level is demanding and is based 37 on DNA sequencing. Determination of HLA alleles is warranted as many HLA alleles are major 38 genetic factors that confer susceptibility to autoimmune diseases and is important for the 39 matching of HLA alleles in transplantation. Here, we compared the accuracy of several 40 published HLA-typing algorithms that are based on next generation sequencing (NGS) data. As 41 genome screens are becoming increasingly routine in research, we wanted to test how well HLA 42 alleles can be deduced from genome screens not designed for HLA typing. The accuracies were 43 assessed using datasets consisting of NGS data produced using the ImmunoSEQ platform, 44 including the full 4 Mbp HLA segment, from 94 stem cell transplantation patients and exome 45 sequences from the 1000 Genomes collection. When used with the default settings none of the 46 methods gave perfect results for all the genes and samples. However, we found that ensemble 47 prediction of the results or modifications of the settings could be used to improve accuracy. Most 48 of the algorithms did not perform very well for the exome-only data. The results indicate that the 49 use of these algorithms for accurate HLA allele determination based on NGS data is not 50 straightforward. 51