HLA-Resolve: Four-field HLA typing and MHC variant detection from long-read hybrid capture

The major histocompatibility complex (MHC) is among the most polymorphic and difficult-to-genotype regions of the human genome. Accurate typing of its HLA genes, which encode antigen-presenting molecules, is critical for transplantation, pharmacogenomics, and disease-risk prediction. Long-read sequencing is increasingly used in HLA typing because of its ability to phase variants across long distances, but many long-read HLA typing workflows still rely on long-range PCR. Here, we present an end-to-end workflow for MHC genotyping that pairs hybrid capture with long-read sequencing on PacBio or Oxford Nanopore, using single-step enzymatic fragmentation and barcoding to enable automated library preparation. The hybrid capture panel enriches both classical and non-classical HLA Class I and Class II genes, as well as all 69 protein-coding MHC Class III genes. We introduce HLA-Resolve, a bioinformatic tool that types HLA genes from PacBio reads via phased full-gene sequence reconstruction, and we benchmark the capture assay and HLA-Resolve together across 32 geographically diverse samples. With PacBio data, variant calling achieved F1 scores of 99.8% for SNVs and 98.4% for indels against the Genome in a Bottle benchmark. HLA-Resolve showed 99.5% concordance with reference HLA typings for the International Histocompatibility Working Group and the Human Pangenome Reference Consortium (HPRC) at three-field resolution and 90.5% at four-field resolution, outperforming three other open-source long-read HLA typers on our HiFi hybrid capture reads. The full-gene sequences reconstructed by HLA-Resolve showed zero edit distance to the corresponding HPRC assemblies in 95% of comparisons. HLA-Resolve also shows high accuracy with whole-genome sequencing (WGS) data, achieving 100% three-field concordance across 40 HPRC PacBio WGS samples. Although developed for the MHC, the workflow is customizable to any gene set, providing a general framework for high-resolution genotyping.