Predicting HLA-DPB1 permissive probabilities through a DPB1 prediction service towards the optimization of HCT donor selection

Despite its demonstrated importance in hematopoietic cell transplantation, the HLA-DPB1 locus is only typed in one in five unrelated donors in the United States. Addressing this issue, we developed a DPB1 Prediction Service that leverages seven-locus haplotype frequencies (HLA-A -C -B -DRB3/4/5 -DRB1 -DQB1 -DPB1) to extend the imputation of six-locus HLA typing (HLA-A -C -B -DRB3/4/5 -DRB1 -DQB1) to the HLA-DPB1 locus, including the novel prediction of HLA-DPB1 TCE groups to calculate donor-recipient TCE permissive match probabilities. Simulations of current-day patient searches reveal the service can fill in missing gaps for another four in five donors that appears on lists. To validate its per-formance, samples of 206,328 registered donors and 5,218 donor-recipient pairs with known high -resolution HLA-DPB1 typing were used for predicted-versus-observed comparisons. These comparisons demonstrated that the predictions were correct for 11.9-19.7% of HLA-DPB1 genotypes, 64.9-70.0% of TCE groups, and 61.0% of permissive match categories. Although HLA-DPB1 match predictions must be confirmed by additional typing, knowledge of TCE match probabilities facilitates rapid and improved identification of best donor options, especially for populations of color. Thus, we developed the TCE Prediction Tool user interface for a pilot program with several transplant centers to preview the accuracy and utility of this prediction framework, which provides valuable upfront optimization of donor selection. (c) 2021 The Authors. Published by Elsevier Inc. on behalf of American Society for Histocompatibility and Immunogenetics. This is an open access article under the CC BY-NC-ND license (http://creativecommons. org/licenses/by-nc-nd/4.0/).

Automated summary

This study introduces a DPB1 Prediction Service to improve donor selection in hematopoietic cell transplantation. By leveraging seven-locus haplotype frequencies, the service can accurately predict HLA-DPB1 genotypes and increase donor-recipient match probabilities. Validation using large sample sizes demonstrates the accuracy and utility of this prediction framework, especially for populations of color.