Please note: Current v1.8.1 not compatible with Tensorflow >= 2.16.0, please downgrade to 2.15.1

pip install tensorflow-cpu==2.15.1

Please see issue here

We are actively working on this, thank you for your understanding.

NanoVar - Structural variant caller using low-depth long-read sequencing

NanoVar is a genomic structural variant (SV) caller that utilizes low-depth long-read sequencing such as Oxford Nanopore Technologies (ONT). It characterizes SVs with using only 4x depth sequencing for homozygous SVs and 8x depth for heterozygous SVs.

Basic capabilities

• Performs long-read mapping (Minimap2) and SV discovery in a single pipeline.
• Accurately characterizes SVs using long sequencing reads (High SV recall and precision in simulation datasets, overall F1 score >0.9)
• Characterizes six classes of SVs including novel-sequence insertion, deletion, inversion, tandem duplication, sequence transposition (TPO) and translocation (TRA).
• Requires 4x and 8x sequencing depth for detecting homozygous and heterozygous SVs respectively.
• Rapid computational speed (Takes <3 hours to map and analyze 12 gigabases datasets (4x) using 24 CPU threads)
• Approximates SV genotype
• Identifies full-length LINE and SINE insertions (Marked by "TE=" in the INFO column of VCF file)
• Repeat element INS annotation using NanoINSight

  Annotating INS variants with NanoINSight

  NanoVar allows the concurrent repeat element annotation of INS variants using NanoINSight.

  To run NanoINSight, simply add "--annotate_ins [species]" when running NanoVar.

  nanovar -t 24 -f hg38 --annotate_ins human sample.bam ref.fa working_dir

  To understand NanoINSight output files, please visit its repository here.

  Installation of NanoINSight dependencies

  NanoINSight requires the installation of MAFFT and RepeatMasker. Please refer to here for instructions on how to install them, or install them through Conda as shown below:

  pip install nanoinsight
  conda install -c bioconda mafft repeatmasker -y

  Note: If encountered "numpy.dtype size changed" tensorflow error while running NanoVar, ensure numpy version is <2.0.0 (i.e. pip install numpy 1.26.4).

  Documentation

  See wiki for more information.

  Versioning

  See CHANGELOG

  Citation

  If you use NanoVar, please cite:

  Tham, CY., Tirado-Magallanes, R., Goh, Y. et al. NanoVar: accurate characterization of patients’ genomic structural variants using low-depth nanopore sequencing. Genome Biol. 21, 56 (2020). https://doi.org/10.1186/s13059-020-01968-7

  Authors

  • Tham Cheng Yong - cytham
  • Roberto Tirado Magallanes - rtmag
  • Asmaa Samy - AsmaaSamyMohamedMahmoud
  • Touati Benoukraf - benoukraflab

  License

  This project is licensed under GNU General Public License - see LICENSE.txt for details.

  Simulation datasets and scripts used in the manuscript

  SV simulation datasets used in the manuscript can be downloaded here. Scripts used for simulation dataset generation and tool performance comparison are available here.

  Although NanoVar is provided with a universal model and threshold score, instructions required for building a custom neural-network model is available here.

  Limitations

  • The inaccurate basecalling of large homopolymer or low complexity DNA regions may result in the false determination of deletion SVs. We advise the use of up-to-date ONT basecallers such as Dorado to minimize this possibility.

  • For BND SVs, NanoVar is unable to calculate the actual number of SV-opposing reads (normal reads) at the novel adjacency as there are two breakends from distant locations. It is not clear whether the novel adjacency is derived from both or either breakends in cases of balanced and unbalanced variants, and therefore it is not possible to know which breakend location(s) to consider for counting normal reads. Currently, NanoVar approximates the normal read count by the minimum count from either breakend location. Although this helps in capturing unbalanced BNDs, it might lead to some false positives.