New tool improves detection of big genome changes from Nanopore data

Detecting large genomic rearrangements (LGRs) — big pieces of DNA that have moved, been duplicated or deleted — is important for understanding both human genetics and infectious organisms. Short-read sequencing technologies have limited ability to resolve these large changes because the reads are too short to span complex events. Long-read sequencing, such as Oxford Nanopore sequencing, produces much longer reads and so promises better detection of LGRs, but the software that calls these events is still evolving and needs optimization. In this work led by corresponding author Andrey Kechin, researchers developed a new program called eLaRodON to identify LGRs specifically from Oxford Nanopore sequencing data. The team tested eLaRodON on publicly available datasets, including Mycobacterium tuberculosis genomes that have extensive complete genome references and the well-studied human cell line NA12878, and compared its performance to existing tools. The program is open source and available on GitHub (https://github.com/aakechin/eLaRodON/), so other groups can examine and use the code as they explore structural variation in bacteria and human genomes.

To evaluate eLaRodON, the researchers used Oxford Nanopore sequencing datasets and compared calls from eLaRodON to those made by NanoSV, Sniffles2, NanoVar, and SVIM. Performance was measured against gold-standard LGR sets that were derived from de novo Flye assemblies and Mauve alignments. On Mycobacterium tuberculosis data, eLaRodON achieved an AUC of 0.61, outperforming the other tools which ranged from 0.13 to 0.43. On the human NA12878 datasets, eLaRodON reached an AUC of 0.86, compared with 0.40–0.72 for the alternatives. The study also revealed recurrent false-positive LGR patterns that appeared across diverse datasets, including M. tuberculosis, NA12878, and λ phage controls, highlighting common pitfalls in long-read LGR calling. Importantly, orthogonal validation by targeted NGS and Sanger sequencing confirmed variants at rates between 67–100% for different LGR types, even for some events supported by a single read, supporting the practical accuracy of eLaRodON.

These results represent a step forward in detecting large genomic rearrangements from Oxford Nanopore data. By improving accuracy over existing callers and providing open access to the code, eLaRodON can help researchers more reliably map structural variation in both pathogens like Mycobacterium tuberculosis and in human genomes such as NA12878. The identification of recurrent false-positive patterns also provides a useful warning: even with improved tools, careful validation using independent methods such as targeted NGS and Sanger sequencing remains important. Because the tool was validated against rigorous gold-standard sets from Flye assemblies and Mauve alignments, laboratories can have greater confidence in the calls it produces, while still applying orthogonal checks for critical findings. Overall, eLaRodON’s combination of improved detection metrics, validation evidence, and open-source availability offers a practical advance for genomic research and potential clinical applications that rely on accurate structural variant detection.