Detecting pathogenic structural variation in families with undiagnosed rare disease in a national genome project.

Detection of de novo structural variants — an AUTS2 inversion, a DLX5/DLX6 locus inversion and an FN1 deletion — by long-read sequencing, confirmed by orthogonal re-analysis of short-read data.

Whole-genome sequencing (WGS) programmes in rare disease reach a diagnostic rate of 25% to 41% depending on patient selection and the extent of prior testing; within the Scottish Genomes Partnership, short-read sequencing had achieved 23% in affected families. The authors applied Oxford Nanopore long-read sequencing to 24 families (74 individuals) that remained undiagnosed after short-read SNV/indel analysis, and retrospectively reviewed structural variant calls derived from the short-read data. After quality, rarity, panel-based and inheritance-based filtering, 392 candidate structural variants were retained across de novo, homozygous recessive, compound heterozygous and X-linked models, of which 8 overlapped PanelApp diagnostic-grade green genes. Pathogenic or likely pathogenic de novo structural variants were identified in 3 of 24 families: an AUTS2 inversion, a DLX5/DLX6 locus inversion and an FN1 deletion. All three were independently confirmed by retrospective short-read re-analysis using SVRare.

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

The most instructive result is a negative one: all three structural variants were present in the short-read data and had not been reported. The limiting factor is therefore not the technology but structural variant calling and prioritisation, which argues for systematic re-analysis of existing data before any substantially more expensive long-read re-sequencing. Across 24 families, a yield of 3 in 24 remains a very imprecise estimate, and selecting families after short-read failure precludes transposing it as such to an unselected cohort.

Analysis by Dr Thibaut Benquey