SVkhor: a unified framework for structural variant integration across long-read, short-read, and optical genome mapping data

Caller-aware normalisation, within-technology merging then cross-technology integration of structural variant calls from short-read sequencing, long-read sequencing and optical genome mapping, with source annotation.

Multi-technology structural variant discovery is hampered by differences in breakpoint resolution, allele representation, annotation and VCF structure across callers and platforms. The authors present SVkhor, a software framework that merges the outputs of multiple callers within each technology and then integrates callsets from short-read sequencing, long-read sequencing and optical genome mapping. It relies on caller-aware normalisation, within-technology merging and cross-technology integration, producing compact, source-annotated structural variant catalogues suitable for benchmarking and downstream interpretation. Evaluation on the HG002 reference sample and analysis of a clinical trio show that SVkhor reduces redundant caller-level complexity while preserving technology-specific evidence, enabling the move from heterogeneous callsets to interpretable sample- and family-level catalogues.

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

The problem addressed is the one facing any laboratory that has stacked short-read, long-read and optical genome mapping: three partial truths and no single catalogue a biologist can report from. SVkhor answers a genuine organisational need, but validation remains thin, HG002 and a single trio, with no sensitivity and specificity metrics by structural variant class and no comparison with existing merging tools, which for now prevents judging what is gained and what is lost by merging. Test it on your own data before any move into production, all the more so as this is a preprint.

Analysis by Dr Thibaut Benquey