Svirlpool: structural variant detection from long read sequencing by local assembly.

Multi-sample structural variant caller (ONT) building local consensus assemblies and retaining assembled sequence through to final joint-calling

Long-read sequencing, particularly Oxford Nanopore, has greatly improved structural variant (SV) detection. Fast alignment-based callers perform well but reduce read sequence to alignment-derived signals, limiting cohort and clinical analyses, especially for insertions and repeat regions. The authors present Svirlpool, a multi-sample caller for ONT data that builds local consensus assemblies of candidate regions and retains the assembled sequence through to final joint-calling, where merging tolerances are scaled by a reference-independent noise estimate. It is validated on two ONT family datasets (HG002 trio, Platinum Pedigree).

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

An original approach — local assembly preserving sequence — well suited to insertions and repeat regions where alignment-based callers struggle, and designed for multi-sample analyses (cohorts, families). A preprint to be confirmed by peer review and on real clinical cohorts, but the 'joint-calling on assembled sequence' angle is relevant to diagnostics.

Analysis by Dr Thibaut Benquey