Structural variant discovery and diagnostic impact in rare diseases from short-read and long-read sequencing

Comparison of structural variant diagnostic yield between short-read and long-read genome sequencing

Across 1,462 families (3,450 individuals) from the Broad CMG and GREGoR programs, short-read genome sequencing with sensitive structural variant detection resolved 5.4% of cases (79/1,462), of which 80% were undetectable by standard cytogenetics. For 96 families, long-read genome sequencing with methylation profiling and long-read transcriptomics was added: it generated over 25,000 structural variants per genome (63% not seen by short-read) but yielded only one additional diagnosis (a mosaic de novo SNV in CASK), an added yield of 1.04%. No diagnosis was provided by long-read transcriptomics or episignatures. A systematic review confirmed that most reported long-read diagnoses were detectable by short-read.

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

This preprint offers a welcome counterpoint to long-read enthusiasm: in a well pre-screened cohort, the additional diagnostic yield of long-read is only 1%, as most of the signal is already captured by short-read coupled with sensitive SV detection. The message is not that long-read is useless, but that its contribution depends critically on the still-nascent annotation of repetitive and noncoding regions. A clear-eyed read to counter overly optimistic estimates, bearing in mind the not-yet-peer-reviewed preprint status.