A high-resolution human pangenome structural variant resource for improved disease association

Population-level structural variant frequency callset built from long-read diploid assemblies, used as a filter for common variation.

From 293 nearly complete long-read diploid genome assemblies, the authors show that while 99% of variants between any two genomes are single base-pair substitutions, 88% of euchromatic variant base pairs are structural variants — insertions, deletions, duplications and inversions. They identify 24 gene-rich regions subject to megabase-scale variation, 2,293 potentially unstable tandem repeats and 890 novel expression quantitative trait loci associated with structural variants. Expanding to 1,218 long-read samples from the 1000 Genomes Project and applying a newly developed cross-platform breakpoint evaluation tool, BoostSV, they build a nonredundant callset of 614,522 structural variants. This callset filters more than 99% of common variation in 44 unsolved long-read probands from the Undiagnosed Diseases Network to isolate likely disease-causing structural variants, and genotyping 1,053 high-impact biallelic structural variants in 232,090 All of Us participants yields 105 significant associations, including 26% where the structural variant is the lead variant.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The useful message for a laboratory is not the size of the callset but the gap between 99% and 88%: our short-read pipelines count variants correctly while missing most of the variant base pairs in the genome. A public population-level structural variant frequency resource is exactly what was missing to turn a long-read call into an interpretable variant — the bottleneck for long-read in diagnostics was never detection, but filtering common variation. Preprint: the association figures and the exact content of the callset await peer review.

Analysis by Dr Thibaut Benquey