ARC-SV is a structural variant caller for paired-end, whole-genome sequencing data. For methodological details, please see our preprint: [https://doi.org/10.1101/200170].

This software was developed in the Wong Lab at Stanford University with funding from the NSF Grant DGE-114747 and NIH grants T32-GM096982, P50-HG007735, and R01-HG007834.

• Requirements
• Installation
  • Example installation from scratch
  • Getting reference resources
• Usage
  • Calling SVs
  • Running the example
  • Filtering and merging output files
  • Description of output

• python 3 (tested with 3.3.2 and 3.5.0)
• python packages (automatically installed by pip)
  • matplotlib
  • numpy
  • pyinter
  • pysam
  • python-igraph
  • scikit-learn
  • scipy

ARC-SV and its dependencies can be installed as follows:

git clone https://github.com/jgarthur/arcsv.git
cd arcsv
pip3 install --user .

OS X users with a brewed Python installation should ignore --user above.

The installed location of the main script, arcsv, must be in your path. The correct folder is probably /usr/bin, /usr/local/bin, or ~/.local/bin.

The following commands should install ARC-SV and all dependencies on a fresh copy of Ubuntu:

# update packages
sudo apt-get update

# install pip and setuptools
sudo apt install python3-pip
pip3 install -U pip setuptools

# extra requirements needed for igraph
sudo apt install libxml2-dev zlib1g-dev

# arcsv setup
sudo apt install git
git clone https://github.com/jgarthur/arcsv.git
cd arcsv
pip3 install --user .

# add this to your .bash_profile
export PATH="~/.local/bin/:$PATH"

You will need a bed file containing the locations of assembly gaps in your reference genome. The resources/ folder contains files for hg19, GRCh37, and hg38, which were retrieved as follows:

curl http://hgdownload.cse.ucsc.edu/goldenpath/hg19/database/gap.txt.gz | \
     zcat | \
     cut -f2-4 > hg19_gap.bed
     
curl http://hgdownload.cse.ucsc.edu/goldenpath/hg38/database/gap.txt.gz | \
     zcat | \
     cut -f2-4 > hg38_gap.bed

or for the NCBI reference (with "2" instead of "chr2"):

curl http://hgdownload.cse.ucsc.edu/goldenpath/hg19/database/gap.txt.gz | \
     zcat | \
     cut -f2-4 | \
     sed 's/^chr//' \
     > GRCh37_gap.bed

To call SVs:

arcsv call -i reads.bam -r chrom[:start-end] -R reference.fasta -G reference_gaps.bed -o output_dir

# To see more detailed documentation on all possible arguments
arcsv call -h

The folder example/ in this repository contains files to test the ARC-SV installation:

arcsv call -i example/input.bam -r 20:0-250000 -o my_example_output \
  -R example/reference.fa -G example/gaps.bed
  
diff my_example_output/arcsv_out.tab example/expected_output.tab

ARC-SV works on a single chromosome at a time. Supposing your output folders are named "arcsv_chr#", you can merge and/or filter the results as follows:

# Recommended settings
arcsv filter-merge --minsize 50 --noinsertions arcsv_chr*

# If no filtering is desired
arcsv filter-merge arcsv_chr*

For each cluster of candidate breakpoints, ARC-SV attempts to resolve the local structure of both haplotypes. The output file arcsv_out.tab contains one line for each non-reference haploype called. A call typically consists of a single SV (simple or complex), but some contain multiple variants that were called together.

Where multiple values are given, as in svtype, the order is left to right in the alternate haplotype, which is shown in the rearrangement column.

All genomic positions in arcsv_out.tab are 0-indexed for compatibility with BED files. (arcsv_out.vcf is still 1-indexed as required.)