Performance of an allele‐level multi‐locus HLA genotype imputation tool in hematopoietic stem cell donors from Quebec

Abstract

IntroductionDonor‐recipient HLA compatibility is an important determinant of transplant outcomes. Allele‐group to allele‐level imputations help assign HLA genotypes when allele‐level genotypes are not available during donor selection.MethodsWe evaluated the performance of HaploStats, an allele‐level multi‐locus HLA genotype imputation tool from the National Marrow Donor Program, in a cross‐sectional study including hematopoietic stem cell donors (HSCD) from Quebec, Canada. A total of 144 self‐identified Caucasian HSCD genotyped at the allele‐group and allele‐level for HLA‐A, ‐B, ‐C, ‐DRB1, and ‐DQB1 loci were studied. We compared allele‐level genotypes imputed by HaploStats to those obtained by the reference standard, sequenced‐based typing (SBT).ResultsImputation performance, determined by allele‐level genotype recall (fraction of matching imputed and sequenced genotypes) was 97%, 96%, 95%, 84%, and 81% for HLA‐A, ‐B, ‐C, ‐DRB1, and ‐DQB1 loci, respectively. Our sample deviated from Hardy‐Weinberg equilibrium only at the HLA‐DRB1 locus. Residual ambiguity, determined by typing resolution scores (TRS), was greatest for HLA class II loci (average TRS 0.65 and 0.80 for DRB1 and DQB1, respectively). In contrast, average TRS of 0.88, 0.84, and 0.92 was observed for HLA‐A, ‐B, and ‐C, respectively.ConclusionsHLA allele imputation from ambiguous genotypes demonstrate satisfactory prediction accuracy for HLA class I but modest prediction accuracy for HLA class II loci in self‐identified Caucasian HSCD from Quebec. While consideration of high‐resolution allele and haplotype frequencies in the Quebec population may improve the performance of available allele‐level multi‐locus genotype imputation tools in Quebec, this study suggests that genotyping at the first two field level should be conducted whenever possible.

Inaccuracies in epitope repertoire estimations when using multilocus allele‐level HLA genotype imputation tools

“…However, inaccurate allele‐level HLA genotype imputation may compromise our ability to verify clinically relevant B‐ and T‐cell epitopes because of addition or omission of epitopes from the true repertoire of donor HLA epitopes. We have previously shown that prediction of allele‐level HLA genotypes by HaploStats show >94% recall for class I but only modest accuracy for class II loci (HLA‐DRB1 and ‐DQB1 recall of 84.03% and 81.25%, respectively) in 144 self‐identified Caucasian HSC donors from Québec . Our findings in the current analysis (recall of 82% and 81% for HLA‐DRB1 and ‐DQB1 loci, respectively) in HSC donor‐recipient pairs reinforce our prior findings.…”

“…HaploStats is a web‐based application provided by the US National Marrow Donor Program Bioinformatics Group that assigns most likely allele‐level HLA types considering haplotype frequencies in the United States whether self‐reported race/ethnicity is available or not (http://www.haplostats.org). A work flow diagram outlining the steps from allele‐group level dataset creation to imputation is available in Table S2. Recall, the fraction of correctly imputed HLA types (ie, matching measured HLA types) from ambiguous allele‐group HLA types, was used to evaluate the accuracy of imputed multilocus allele‐level HLA genotypes of HSC donors from both study samples.…”

“…To inform on donor‐recipient epitope compatibility, unambiguous, allele‐level HLA genotypes are required. Allele‐level HLA types are only available in a minority of donors, mainly because of time constraints in solid organ transplantation and added costs in comparison to allele‐group‐level HLA types. To overcome these restrictions, the repertoire of HLA epitopes can be estimated from allele‐level HLA types imputed from allele‐group HLA types.…”

Substituting imputation of HLA antigens for high-resolution HLA typing: Evaluation of a multiethnic population and implications for clinical decision making in transplantation

“…Unfortunately, as we have demonstrated, this approach introduces a significant level of error. Our results, showing an accuracy of 45.2% for imputation of a 10‐allele HLA type among non‐Hispanic Caucasians, are similar to the 46% accuracy reported by Dr Sapir‐Pichhadze's team in an exclusively Caucasian population 11 . Importantly, accuracy was significantly worse for non‐Caucasians.…”

Human leukocyte antigen eplet mismatches and long‐term clinical outcomes in pediatric renal transplantation: A pragmatic, registry‐based study

“…We used the available HLA typing in combination with information on local haplotype frequencies and extended class II linkage associations to determine the most likely complete, high‐resolution typing. There is limited evidence that using haplotype assumed based on high‐resolution typing to determine quantitative EpMM may be sufficiently accurate for the purpose of epidemiological studies 34,35 ; however, due to this major limitation, our findings should be viewed as hypothesis generating only.…”