90
P050
Abstracts / Human Immunology 78 (2017) 51–254
RAPID, HIGH RESOLUTION HLA GENOTYPING USING NANOPORE SEQUENCING Peter M. Clark a, Deborah Ferriola a, Dimitri S. Monos a,b. aChildren’s Hospital of Philadelphia, Philadelphia, PA, United States; bUniversity of Pennsylvania Perelman School of Medicine, Philadelphia, PA, United States. Aim: Despite the significant advancement of high resolution HLA genotyping by next generation sequencing (NGS), the full promise of NGS has yet to be fully realized. Paired-end Illumina sequencing of HLA amplicons results in lengthy turn around times often approaching 72 h and remains unable to resolve genotyping ambiguities within class II HLA genes. We aim to address both of these issues by performing rapid, single molecule sequencing of full length HLA amplicons on the Oxford Nanopore sequencing platform. Methods: HLA amplicons were generated at 11 loci (HLA-A, HLA-B, HLA-C, HLA-DQA1, HLA-DQB1, HLADPA1, HLA-DPB1, HLA-DRB1, HLA-DRB3, HLA-DRB4 and HLA-DRB5) using the Omixon Holotype kit for five individuals with ambiguous HLA-DPB1 genotyping (as assessed by the analysis of paired-end sequencing data). HLA amplicons were used to generate 2D sequencing libraries that were sequenced on the Oxford Nanopore MinION platform with real-time base calling. A custom bioinformatics pipeline was developed to generate phased, full-length HLA allele consensus sequences and perform high resolution genotyping for each amplified locus. Results: Sequencing reads were found to have a high insertion/deletion error rate (particulary around homopolymer stretches) and an observed average basecall error rate of 4%. However, with an average read depth greater than 200X per locus, our analytical pipeline was able to generate highly accurate (99.99%), fulllength consensus sequences and unambiguous HLA genotyping results for each HLA locus analyzed that were 100% consistent with genotyping results obtained from the analysis of paired-end sequencing data. All steps of the process, from DNA extraction to high resolution HLA genotyping for each sample was performed within 24 h. Conclusions: Our results demonstrate the utility of the Oxford Nanopore MinION sequencing platform in combination with our developed bioinformatics pipeline to generate full-length gene consensus sequences and unambiguous high resolution HLA genotyping results. As the platform chemistry and flow-cell continue to be optimized, we hope that turn around time and read accuracy will further improve, facilitating unambiguous, rapid high resolution HLA genotyping within Immunogenetics laboratories addressing currently unmet needs.
P051
A MACHINE LEARNING MODEL THAT PREDICTS BINDING BETWEEN KIR3DL1 AND HLA CLASS I ALLOTYPES Martin Maiers a, Yoram Louzoun b, Phill Pymm c, Julian Vivian c, Jamie Rossjohn c, Andrew Brooks d. aNational Marrow Donor Program, Minneapolis, MN, United States; bBar-Ilan University, Ramat Gan, Israel; cMonash University, Clayton, Australia; dUniversity of Melbourne, Parkville, Australia. Aim: KIR3DL1 is a polymorphic inhibitory NK receptor that recognizes HLA class I allotypes that contain the Bw4 motif. Structural analyses have shown that in addition to residues 77–83 that span the Bw4 motif, polymorphism at other sites throughout the HLA molecule can influence the interaction with KIR3DL1. Given the extensive polymorphism of both the KIR3DL1 and HLA class I we sought to train and evaluate a model for predicting the binding between any combination of KIR3DL1 and HLA class I allotypes. Methods: KIR3DL1 tetramers were screened for reactivity against a panel of HLA-I molecules which revealed different patterns of specificity for each KIR3DL1 allotype. This data was used to train several machine learning models (Support Vector Machine, Multi-Label Vector Optimization, Linear Regression, Neural Network) to learn the association between the amino acid sequence of the HLA class I allotype and the normalized MFI of the tetramer binding experiments. Separate models were trained for each of 6 KIR3DL1 allotypes. Results: The performance of the predictor was evaluated by random cross-fold (80/20) validation and computing the area under the curve (AUC) of the receiver-operator characteristic (ROC). The MLVO model performed best with AUC scores ranging from 0.795 to 0.846 for the 6 KIR3DL1 allotype models. Conclusions: This binding predictor can be applied to clinical datasets to more specifically investigate the role of HLA-KIR interaction and improve clinical decision making.