Benchmarking long-read variant sensitivity across ONT and PacBio platforms using known clinically reported variants in a cohort of critically ill newborns.

Retrospective comparison of variants reported by clinical short-read WGS with those identified by long-read WGS on the Oxford Nanopore and PacBio platforms in the same families.

Long-read WGS is presented as an all-in-one test able to detect clinically relevant variants, including those missed by current short-read pipelines, yet its sensitivity for variants already identified and prioritised by clinical short-read WGS had not been assessed. Within the SeqFirst-neo study, a subset of critically ill newborns and their parents who underwent clinical short-read WGS were also sequenced with long-read WGS on the Oxford Nanopore and PacBio platforms: 134 families in total, including 128 with clinical short-read WGS sequenced on both long-read platforms. Among these 128 families, 89 SNV/indels and 14 structural variants or CNVs reported by the clinical short-read pipeline were evaluated: all variants assessed in probands were ultimately detected by both long-read platforms, although three events were not detected before an updated variant caller was applied. Breakpoint coordinates and event sizes often differed substantially between short-read and long-read calls, leaving persistent challenges for direct comparison, particularly for structural variants and CNVs.

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

This is the reassuring result awaited by laboratories considering a move to long-read sequencing: nothing reported by short-read testing is lost, across 128 families and two platforms. Two serious caveats remain: three events were only recovered after the variant caller was updated, meaning performance depends as much on software as on chemistry and dates quickly; and discrepancies in breakpoints and event sizes make structural variant and CNV comparison delicate, with a practical risk of wrongly concluding discordance. The study measures retrospective sensitivity on what was already known, not the added diagnostic value of long-read sequencing — that is the next question, and it remains open.

Analysis by Dr Thibaut Benquey