Abstract 5116: Benchmarking the HLA typing performance of three HLA assay and two software

Abstract

Abstract Background:The adaptive immune response intrinsically depends on hyper variable human leukocyte antigen (HLA) genes. Correct HLA typing is crucial for successful donor-patient matching in organ transplantation and recently is also an emerging biomarker for immune checkpoint inhibitor. With the great progress on next generation sequencing (NGS) technology, sequencing accuracy and throughput have increased while the cost for data have decreased faster than Moore's Law, thus HLA tying by NGS has been widely used from laboratory to clinic. Method:We tested three NGS assay for HLA typing and two recently published HLA typing software from 24 samples. Three assays involve two target capture methods (WES and HLA target sequencing) and one amplification sequencing method. Two software involve Xhla and HLAHD, respectively. First, we compared 3 assay performance based on PE150 read length and xHLA software by both 4 alleles and 8 alleles. Second, we compared the performance of two widespread HLA typing software xHLA and HLA-HD according to different assay type, read depth, read length and speed. Results:First of all, for PE150 datasets on three assay, the accuracy rate of HLAHD is all 100%, while xhla in WES is 99%, amplicom 98.6%, target HLA sequencing 100%. On reads length evaluation, the accuracy of HLAHD decreases gradually from PE150 to PE100 and PE75, and the capture method is more sensitive to reads length than the amplification assay, while Xhla is very robust on read length. On coverage evaluation, when the amplicom sequencing data was down-sampled 500X to 10X, the accuracy of HLAHHD and XHLA is both 100% in 50X and above and decrease gradually when less than 50X. In terms of time, with the increase in sequencing depth, running time for both software increase. HLAHD prediction from fastq file is longer than that from bam file, especially when the panel data is relatively large, the time will increase exponentially. Conclusions:Our benchmarking dataset show that current HLA assay and computational methods for HLA typing generally provide satisfactory results, while different assay type matched with different software, read depth and read length do affect the performance. Citation Format: Rui Meng, Tao Zhu, Yu Gong, Yanan Chen, Hua Dong, Ye Yizhou, Fugen Li, Ping Liu, Minya Yao. Benchmarking the HLA typing performance of three HLA assay and two software [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 5116.