A diploid assembly-based benchmark for variants in the major histocompatibility complex.

Area of Science:

• Genomics
• Human Genetics
• Bioinformatics

Background:

• Human genome analysis traditionally relies on mapping reads to a reference genome.
• Accurate, phased de novo assemblies are now achievable with advanced long-read and linked-read sequencing technologies.
• The Major Histocompatibility Complex (MHC) is a medically significant, highly variable genomic region where diploid assembly is crucial.

Purpose of the Study:

• To develop a high-quality, phased diploid genome benchmark for the human MHC region.
• To assess the performance of variant calling methods in complex genomic areas.
• To create a more comprehensive variant benchmark for the MHC than previously available.

Main Methods:

• Utilized long-read and linked-read sequencing data for the Genome in a Bottle sample HG002.
• Performed phased de novo assembly to generate separate contigs for each haplotype.
• Aligned haplotype assemblies to the reference genome and called phased small and structural variants.

Main Results:

• Generated a single contig for each haplotype of the HG002 sample.
• Defined a small variant benchmark for the MHC, covering 94% of the region and including 22,368 variants (<50 bp).
• The new benchmark identified 49% more variants than a mapping-based approach and revealed errors in existing callsets.

Conclusions:

• Phased de novo assembly provides a superior approach for characterizing complex genomic regions like the MHC.
• The developed MHC benchmark enhances the accuracy of variant detection and performance assessment.
• This work facilitates better understanding and analysis of medically important, highly variable genomic regions.