Pangenome-based structural variant discovery

SVarp discovers haplotype-resolved structural variants (SVs) on pangenome graphs using long-read sequencing data. It outputs local assemblies of SV alleles (svtigs).

For questions, please open an issue.

conda create -n svarp -c conda-forge -c bioconda svarp
conda activate svarp
svarp --help

git clone https://github.com/asylvz/SVarp.git
cd SVarp

SVarp depends on:

• zlib (sudo apt-get install zlib1g-dev)
• GraphAligner (https://github.com/maickrau/GraphAligner)
• Samtools (https://github.com/samtools/samtools)
• HTSlib (https://github.com/samtools/htslib/)
• WFA2-lib (https://github.com/smarco/WFA2-lib)
• Minimap2 (https://github.com/lh3/minimap2)

Note: wtdbg2, Minimap2, and Samtools are bundled under third_party/ and built automatically. Their paths are resolved relative to the svarp executable, so the binary is portable.

You can download and build HTSlib, WFA2-lib, wtdbg2, Minimap2, Samtools using:

make libs

SVarp is developed and tested on Linux. It compiles and runs on macOS ARM, but has not been fully tested due to GraphAligner lacking native ARM support. gcc is required to build the bundled wtdbg2 (brew install gcc).

make

Binary will be at build/svarp.

ONT (default):

build/svarp \
    --gaf sample.gaf \
    --graph pangenome.gfa \
    --fasta reads.fasta.gz \
    --sample SAMPLE1 \
    --out output_dir

PacBio HiFi:

build/svarp \
    --gaf sample.gaf \
    --graph pangenome.gfa \
    --fasta reads.fasta.gz \
    --reads hifi \
    --sample SAMPLE1 \
    --out output_dir

• Reads FASTA must be bgzipped (i.e., bgzip NA12878.fasta)
• GAF must follow the unstable rGFA coordinate system (https://github.com/lh3/gfatools/blob/master/doc/rGFA.md)
• Graph must be provided in GFA format. We tested SVarp with HPRC Minigraph pangenome (https://zenodo.org/record/6983934)
  • GAF alignments must be generated using the same GFA file that you use as input
• Phasing tags (optional) should be generated using WhatsHap: (https://whatshap.readthedocs.io/)

whatshap haplotag input.vcf.gz input.bam \
    --reference ref.fasta \
    --output-haplotag-list tags.tsv \
    -o phased.bam

build/svarp \
    --gaf sample.gaf \
    --graph pangenome.gfa \
    --fasta reads.fasta.gz \
    --reads hifi \
    --phase read_tags.tsv \
    --sample SAMPLE1 \
    --out output_dir

Required arguments:
--gaf (-a)                  : Alignment file in GAF format
--graph (-g)                : Pangenome file in GFA format
--fasta (-f)                : Fasta sequence file


Optional arguments:
--sample (-i)               : Sample name.
--out (-o)                  : Output folder.
--debug                     : Output multiple log files for debugging purpose.
--skip-untagged             : Output only phased variants (~30% faster).
--dist-threshold (-d)       : Distance threshold to merge SV breakpoints (default=100)
--phase (-p)                : WhatsHap haplotag file in .tsv (https://whatshap.readthedocs.io/en/latest/guide.html#whatshap-haplotag)
--reads (-w)                : Read type: ont (default), hifi, or clr. Sets wtdbg2 preset and assembly parameters.
--support (-s)              : Minimum support for a cluster to be assembled (default=5 for diploid samples)
--threads (-t)              : Number of threads for assembly and realignment (default=16)
--version (-v)              : Print version
--help (-h)                 : Print this help menu

Soylev, A., Ebler, J., Pani, S., Rausch, T., Korbel, J., & Marschall, T. (2024). SVarp: pangenome-based structural variant discovery. bioRxiv, 2024-02.