Multi-genome benchmark for simple and complex genome inversion detection

Benchmark set of simple and complex inversions derived from Strand-seq and phased long-read assemblies across five reference samples

Inversions are structural variants involved in genomic disorders among other roles, but their detection remains challenging in repeat-dense regions and for complex multi-breakpoint events. The authors present a benchmark of simple and complex inversions across five reference samples, derived from Strand-seq and phased long-read assemblies, with breakpoint refinement using haplotype-resolved assemblies. Across short-read, PacBio HiFi and Oxford Nanopore data, callers recovered simple inversions with high sensitivity at sufficient coverage, whereas complex and heterozygous events remained difficult. Sniffles2 and Severus achieved the strongest recall for complex inversions despite more false positives, and the choice of mapper (Minimap2 or VACmap) substantially affected detection in repetitive regions.

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

Two useful messages for a laboratory: a complex or heterozygous inversion remains hard to detect whatever the platform, and the mapper matters as much as the caller in repetitive regions. The benchmark relies on five reference genomes, with no patient data and no sensitivity figures in the abstract: it is a basis for tuning a long-read pipeline, not a diagnostic validation.

Analysis by Dr Thibaut Benquey