Deriving HLA genotyping from whole genome sequencing data using Omixon HLA twin(tm) in G3’s global clinical study

Abstract

Aim Identify the degree of concordance and the ambiguities when DNA samples are analyzed using two HLA genotyping systems; Whole Genome Sequencing (WGS) data analyzed by Omixon HLA Target software and genotyping by the Holotype HLA typing system. Methods WGS data were generated of 235 human samples from G3’s GLOBAL study (NCT01738828) using paired 100 bps long HiSeq Illumina ∼30× coverage and analyzed by the Omixon HLA Target software. Genotyping was also performed for the same samples using the Holotype HLA typing system, whereby the generated libraries were sequenced on the Illumina MiSeq using the 250 bps paired-end sequencing protocol. Data were analyzed by the Omixon HLA Twin software. Concordance at the three-field level between the two approaches was assessed for each of the HLA-A, B, C, DQB1 and DRB1 loci. Ambiguities were also determined for each of the typing methods. Results A pairwise comparison between the two NGS-based methods shows that in average they were 97.2% concordant at the three-fields level (HLA-A: 450/470 95.7%, HLA-B: 464/470 98.7%, HLA-C: 464/470 98.7%, HLA-DQB1: 438/470 93.2%, HLA-DRB1 467/468 99.8%). Most of the discrepancies were due to allele dropout in the WGS sample or systematic errors of the software. The 41 non-systematic mistypings were due to either insufficient coverage of the WGS data or algorithmic problems. The Holotype HLA system did not generate any ambiguities for the HLA-A, -B, -C and HLA-DQB1 loci at the three-fields typing level. However, as intron 1 and exon 1 of DRB1 are not targeted, 1.4% of ambiguities were generated for the DRB1 locus.The WGS typing considers only the exons of the reference database, leading to phase ambiguities and overall ambigous typings of 9.8%, 4.3%, 7.7%, 8.3% and 3.4% for HLA-A, B, C, DQB1 and DRB1 respectively. Conclusions HLA genotyping using WGS data derived through next-generation sequencing and analyzed by the Omixon HLA Twin software is possible at 97.2% accuracy level. Future improvements of the software addressing ambiguities and remaining systematic errors can make HLA typing from WGS data more reliable. S. Juhos: Employee; Company/Organization; Omixon Ltd. T. V a g o : Employee; Company/Organization; Omixon Ltd. D. Ferriola: Employee; Company/Organization; Children’s Hospital of Philadelphia. J. Duke: Employee; Company/Organization; Children’s Hospital of Philadelphia. S. V o r o s: Employee; Company/Organization; Global Genomics Group. B.O. Brown: Employee; Company/Organization; Global Genomics Group. I. Marvasty: Employee; Company/Organization; Global Genomics Group. T. Hague: Employee; Company/Organization; Omixon Ltd. D. Monos: Employee; Company/Organization; Children’s Hospital of Philadelphia.