Comparison of exome-based HLA class I genotyping tools: identification of platform-specific genotyping errors

Our plain-English summary of Comparison of exome-based HLA class I genotyping tools: identification of platform-specific genotyping errors, by K. Kiyotani, T.H. Mai, Y. Nakamura (Journal of Human Genetics, 2017).

The study

• Authors: K. Kiyotani, T.H. Mai, Y. Nakamura
• Journal: Journal of Human Genetics
• Published: 2017;62(3):397-405
• Cited by: 47 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Accurate human leukocyte antigen (HLA) genotyping is critical in studies involving the immune system. Several algorithms to estimate HLA genotypes from whole-exome data were developed. We compared the accuracy of seven algorithms, including Optitype, Polysolver and PHLAT, as well as investigated patterns and possible causes of miscalls using 12 clinical samples and 961 individuals from the 1000 Genomes Project. Optitype showed the highest accuracy of 97.2% for HLA class I alleles at the second field resolution, followed by 94.0% in Polysolver and 85.6% in PHLAT. In Optitype, 34 (21.1%) of 161 miscalls were across different serological types, and common miscalls were HLA-A*26:01 to HLA-A*25:01, HLA-B*45:01 to HLA-B*44:15 and HLA-C*08:02 to HLA-C*05:01 with error rates of 4.1%, 10.0% and 4.1%, respectively. In Polysolver, 193 (55.9%) of 345 miscalls occurred across different serological alleles, and a specific pattern of genotyping error from HLA-A*25:01 to HLA-A*26:01 was observed in 93.3% of HLA-A*25:01 carriers, due to dropping of HLA-A*25:01 sequence reads during the extraction process of HLA reads. In PHLAT, 147 (59.8%) of 246 miscalls in HLA-A were due to erroneous assignment of multiple alleles to either HLA-A*01:22 or HLA-A*01:81. These results suggest that careful considerations needed to be taken when using exome-based HLA class I genotyping data and applying these results in clinical settings.

In plain English

This methods study compared computer programs that work out a person's HLA type (immune-system genes) from whole-exome DNA sequencing data. It did not study asbestos exposure. Its 12 clinical samples came from patients with malignant mesothelioma and were obtained under University of Chicago ethics approval. In those samples, the seven programs' accuracy at second-field (protein-level) resolution ranged from 36.1 to 100% of 72 alleles, checked against laboratory typing. The authors then tested the three best programs on 961 individuals from the 1000 Genomes Project. At second-field resolution, Optitype was correct for 97.2% of alleles, Polysolver for 94.0% and PHLAT for 85.6%. Each program made its own typical mistakes. Polysolver misread HLA-A*25:01 as HLA-A*26:01 in 14 cases, 93.3% of HLA-A*25:01 carriers. The authors traced this to the reference genome used to pick out HLA reads, which carries only one HLA-A allele. PHLAT wrongly assigned various alleles to HLA-A*01:22 or HLA-A*01:81. For Optitype, samples with fewer HLA reads were typed less accurately. The authors concluded that further studies and validation are needed before clinical use.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Methods comparison study of computer tools that estimate HLA class I genotypes from whole-exome sequencing data, checked against laboratory HLA typing
• Who was studied: Genomic DNA from 12 patients with malignant mesothelioma, obtained under University of Chicago Institutional Review Board approval (no dates given), plus exome data from 961 individuals across 13 populations in the 1000 Genomes Project
• Exposure: No asbestos or other environmental exposure was studied. The study compared seven HLA typing programs (OptiType, Polysolver, PHLAT, HLAreporter, HLAforest, HLAminer and seq2HLA).
• Main results:
  • In the 12 clinical samples, accuracy for HLA-A, B and C ranged from 36.1 to 100% across the seven programs. Accuracy was measured as the share of 72 alleles correctly identified at the second field level. The three programs above 90% were chosen for further testing.
  • In the 961 1000 Genomes Project samples, Optitype was accurate for 97.2% of HLA class I alleles at the second field level, Polysolver for 94.0% and PHLAT for 85.6%
  • PHLAT incorrectly assigned nearly 20% of HLA-A alleles
  • In Optitype, 44 of 51 HLA-A errors (86.3%) were within the same serological allele group
  • In Polysolver, the most common HLA-A miscall (14 of 126 HLA-A discordances) was HLA-A*25:01 read as HLA-A*26:01. This error was seen in 93.3% of HLA-A*25:01 carriers.
  • The authors identified the main cause of the Polysolver miscalls as the first step of pulling out HLA reads using the hg19 reference genome, which contains only the HLA-A*03:01 allele
  • PHLAT wrongly assigned various alleles to HLA-A*01:22 or HLA-A*01:81. These errors were specific to PHLAT and were not found in Optitype or Polysolver.
  • For Optitype, accuracy was 82.0% in samples with fewer than 1000 HLA reads (N = 25), 95.4% with 1000 to 2000 reads (N = 193) and 98.0% with more than 2000 reads (N = 743)
• The authors' conclusion: The authors concluded that certain miscall patterns still occurred even with Optitype, and that further studies and validation are needed to improve accuracy before these techniques are applied in clinical settings
• Limitations the authors state:
  • The best tool's accuracy (Optitype, 97.2%) was still lower than PCR-based next-generation sequencing HLA typing, which the authors describe as nearly 100% accurate
  • The number of HLA reads explains only part of Optitype's miscalls
  • The exact causes of some miscalls, such as HLA-C*08:02 read as HLA-C*05:01, remain unclear

Topics

The National Library of Medicine indexes this study under subjects our site covers: Mesothelioma.

Related studies

• Whole exome and targeted deep sequencing identify genome-wide allelic loss and frequent SETDB1 mutations in malignant pleural mesotheliomas — H.C. Kang, H.K. Kim, et al. (2016)
• Driver Gene and Novel Mutations in Asbestos-Exposed Lung Adenocarcinoma and Malignant Mesothelioma Detected by Exome Sequencing — S. Mäki-Nevala, V.K. Sarhadi, et al. (2016)
• Errors and blunders in the estimation of radiation exposure for the uranium miners of WISMUT in Germany — G.G. Eigenwillig (2006)
• Comparison of the ultra-low-dose Veo algorithm with the gold standard filtered back projection for detecting pulmonary asbestos-related conditions: a clinical observational study — M. Tekath, F. Dutheil, et al. (2014)
• Diffusion-weighted MRI of malignant pleural mesothelioma: preliminary assessment of apparent diffusion coefficient in histologic subtypes — R.R. Gill, S. Umeoka, et al. (2010)
• Robust Cell Detection and Segmentation for Image Cytometry Reveal Th17 Cell Heterogeneity — T. Tsujikawa, G. Thibault, et al. (2019)

Browse return to medical research.