A pan-MHC reference graph with 246 fully contiguous phased sequences

Abstract

The major histocompatibility complex (MHC) is a region of the human genome that is key to immune system function but sometimes refractory to genomic analyses due to extreme polymorphism and structural variation. We performed targeted long-read sequencing and de novo assembly of MHC to create 246 highly accurate, fully contiguous, and phased full-length sequences, mostly from data provided by the Human Pangenome Reference Consortium (HPRC). We identified alleles at high resolution across 39 loci including the class I and II HLA (human leukocyte antigen) genes, discovering 1,246 putative novel allele sequences. We identified copy number variation in the C4A and C4B genes and found significant linkage disequilibrium between C4A~C4B haplotypes and 14 MHC loci. We build our sequences into a novel "pan-MHC" reference graph, and we demonstrate that this improves the accuracy of short-read variant calling. Our haplotypes and graph contain significantly more population diversity than preexisting MHC sequences, thus improving the prospects for global health equity in this clinically important genomic region.

MHConstructor: A high-throughput, haplotype-informed solution to the MHC assembly challenge

“…The nucleotide length of each error was summed for each size category and a final percent error was calculated for each assembly by dividing the number of erroneous base pairs in each category by the full length of assembled sequence. To evaluate the impact of heterozygous, diploid sequence data on assembly accuracy, we used three 1000GenomesProject 30X WGS samples with long read data that has recently been assembled into phased MHC haplotypes (Huijse et al . 2023).…”

“…rate. To further validate this method and ensure that it can successfully handle the heterozygous, diploid sequence data found in human populations, we used MHConstructor to re-assemble phased, diploid MHC reference sequences (Huijse et al 2023) from corresponding 1000GenomesProject 30X WGS reads. We generated the assemblies using the long-read sequenced, phased haplotype sequence as the guide sequence.…”