A Tool for the Assessment of HLA-DQ Heterodimer Variation in Hematopoietic Cell Transplantation

When optimizing transplants, clinical decision-makers consider HLA-A, -B, -C, -DRB1 (8 matched alleles out of 8), and sometimes HLA-DQB1 (10 out of 10) matching between the patient and donor. HLA-DQ is a heterodimer formed by the β chain product of HLA-DQB1 and an α chain product of HLA-DQA1. In addition to molecules defined by the parentally inherited cis haplotypes, α-β trans-dimerization is possible between certain alleles, leading to unique molecules and a potential source of mismatched molecules. Recently, researchers uncovered that clinical outcome after HLA-DQB1-mismatched unrelated donor HCT depends on the total number of HLA-DQ molecule mismatches and the specific α-β heterodimer mismatch. Our objective in this study is to develop an automated tool for analyzing HLA-DQ heterodimer data and validating it through numerous datasets and analyses. By doing so, we provide an HLA-DQ heterodimer tool for DQα-DQβ trans-heterodimer evaluation, HLA-DQ imputation, and HLA-DQ-featured source selection to the transplant field. In our study, we leverage 352,148 high-confidence, statistically phased (via a modified expectation-maximization algorithm) HLA-DRB1∼DQA1∼DQB1 haplotypes, 1,052 pedigree-phased HLA-DQA1∼DQB1 haplotypes, and 13,663 historical transplants to characterize HLA-DQ heterodimers data. Using our developed QLASSy (HLA-DQA1 and HLA-DQB1 Heterodimers Assessment) tool, we first assessed the data quality of HLA-DQ heterodimers in our data for trans-dimers, missing HLA-DQA1 typing, and unexpected HLA-DQA1 and HLA-DQB1 combinations. Since trans-dimers enable up to four unique HLA-DQ molecules in individuals, we provide in-silico validations for 99.7% of 275 unique trans-dimers generated by 176,074 U.S. donors with HLA-DQA1 and HLA-DQB1 data. Many individuals lack HLA-DQA1 typing, so we developed and validated high-confidence HLA-DQ annotation imputation via HLA-DRB1 with >99% correct predictions in 23,698 individuals. A select few individuals displayed unexpected HLA-DQ combinations. We revisited the typing of 61 donors with unexpected HLA-DQ combinations based on their HLA-DQA1 and HLA-DQB1 typing and corrected 22 out of 61 (36%) cases of donors through data review or retyping and used imputation to resolve unexpected combinations. After verifying the data quality of our datasets, we analyzed our datasets further: we explored the frequencies of observed HLA-DQ combinations to compare HLA-DQ across populations (for instance, we found more high-risk molecules in Asian/Pacific Islander and Black/African American populations), demonstrated the effect of HLA-DQA1 and HLA-DQB1 mismatching on HLA-DQ molecular mismatches, and highlighted where donor selections could be improved at the time of search for historical transplants with this new HLA-DQ information (where 51.9% of G2-mismatched transplants had lower-risk, G2-matched alternatives). We encapsulated our findings into a tool that imputes missing HLA-DQA1 as needed, annotates HLA-DQ (mis)matches, and highlights other important HLA-DQ data to consider for the present and future. Altogether, these valuable datasets, analyses, and a culminating tool serve as actionable resources to enhance donor selection and improve patient outcomes.