• Why extracting start/end SV breakpoints from a VCF is not trivial
• Requirements
• Quick Installation
  • Install from PyPI
  • Upgrade
  • Install from GitHub
  • Run the test suite
• Command line usage / Options
• Outputs
• Variant filtering rules
  • Behavior
• How to cite?
• Example application: Cohort assessment of SV presence/absence using gnomAD v4 SVs as reference
• License

Why extracting start/end SV breakpoints from a VCF is not trivial

In an SV VCF, the first breakpoint is usually straightforward to retrieve from the CHROM and POS columns. However, the second breakpoint is not encoded in a single standardized way and may appear in different fields depending on the SV type or the caller.

SV representation	First breakpoint	Second breakpoint
Symbolic allele (<DEL>, <DUP>, <INV>, <CNV>)	CHROM:POS	usually INFO/END
Breakend notation (e.g. ]chr13:53040041]ATATATATACACACA)	CHROM:POS	embedded in the ALT field
Sequence notation (e.g. INS: REF=A and ALT=ATGATTCGTTCTG...)	CHROM:POS	embedded in the REF field
Sequence notation (e.g. DEL: REF=TGGAATTAGCCTG... and ALT=T)	CHROM:POS	embedded in the REF field
Caller-specific representations	CHROM:POS	may use alternative tags such as SVEND

As a consequence, extracting both breakpoints from an SV VCF requires handling multiple representations. breakpoint2BedSV addresses this issue by converting heterogeneous SV representations into a unified BED-like breakpoint format.

Requirements

cf pyproject.toml

Quick Installation

Install from PyPI

The recommended way to install breakpoint2BedSV is with pip:

pip install breakpoint2bedsv

Then verify the installation:

breakpoint2bedsv --help

Upgrade

To upgrade to the latest version:

pip install --upgrade breakpoint2bedsv

Install from GitHub

To install the latest development version directly from GitHub:

git clone https://github.com/lgmgeo/breakpoint2BedSV.git
cd breakpoint2BedSV
poetry install

Then run:

poetry run breakpoint2bedsv --help

Run the test suite

To run all tests locally:

poetry run pytest -v

To list the collected tests without executing them:

poetry run pytest --collect-only

The test data and test scripts are located in the tests/ directory.

All tests are also executed automatically through GitHub Actions on each push and pull request.

Command line usage / Options

usage: breakpoint2bedsv [-h] [-V] [--log-file <File>] -i <File> [-d <Dir>] -o <File> [-T <Dir>] [-v]

Convert SV breakpoints from VCF/BCF to BED

optional arguments:
  -h, --help            show this help message and exit
  -V, --version         show program's version number and exit
  --log-file <File>     write log messages to the specified file

Input files:
  -i <File>, --input-file <File>
                        the SV VCF/BCF input file
                        VCF/VCF.gz/BCF files are supported
                        multi-allelic lines are not allowed
                        required

Output options:
  -d <Dir>, --output-dir <Dir>
                        the output directory
                        default: current directory
  -o <File>, --output-file <File>
                        output BED file containing non redundant SV breakpoints
                        (VCF/BCF IDs are merged as a comma-separated list when
                        multiple variants share the same coordinates)
                        required

Behavior:
  -T <Dir>, --tmp-dir <Dir>
                        directory where temporary files will be created
                        if not provided, the system default temporary directory is used
  -v, --verbose         enable verbose output

Outputs

Running the tool will generate a BED output file with SV start/end coordinates and the associated VCF ID. Redundant genomic coordinates are merged into a single BED entry, with multiple VCF IDs reported as a comma-separated list.

Variant filtering rules

breakpoint2BedSV only processes structural variants compatible with breakpoint-based BED representation.

During parsing, the following records are automatically ignored:

• FILTER = MULTIALLELIC (including MCNV-like multi-allelic CNV representations)
• ALT = <BND> (breakend complex rearrangements)
• ALT = <CPX> (complex structural variants)
• ALT = <CTX> (complex translocations)

Behavior

• These variants are skipped during parsing
• They are not written to the output BED file
• The number of skipped records is reported as a warning in the standard output

How to cite?

Please cite the following doi if you are using this tool in your research:

Example application: Cohort assessment of SV presence/absence using gnomAD v4 SVs as reference

Aim

=> Annotate PE/SR-based SVs in a VCF with a gnomAD_excl flag when at least one breakpoint overlaps a gnomAD v4 SV exclusion region. (cf discussion in the gnomAD forum)

Workflow

SV VCF
  │
  ├── breakpoint2BedSV
  │      → convert all SVs into breakpoint-level BED intervals
  │
  ├── bedtools intersect
  │      → overlap SV breakpoints with gnomAD v4 SV exclusion regions
  │
  ├── collect overlapping SV IDs
  │
  └── annotate VCF
         → add INFO flag: gnomAD_excl

GRCh38 gnomAD exclusion resources

SV calling is less reliable in some genomic regions due to:

• low mappability / depth bias
• peri-centromeric or peri-telomeric repeats
• known problematic regions in population datasets such as gnomAD

Two GRCh38 gnomAD exclusion regions:

• depth_blacklist.sorted.bed.gz
• PESR.encode.peri_all.repeats.delly.hg38.blacklist.sorted.bed.gz

curl -O https://storage.googleapis.com/gatk-sv-resources-public/hg38/v0/sv-resources/resources/v1/depth_blacklist.sorted.bed.gz
curl -O  https://storage.googleapis.com/gatk-sv-resources-public/hg38/v0/sv-resources/resources/v1/PESR.encode.peri_all.repeats.delly.hg38.blacklist.sorted.bed.gz

Output

SVs with at least one breakpoint overlapping one of these exclusion regions are flagged in the VCF with:

##INFO=<ID=gnomAD_excl,Number=0,Type=Flag,Description="At least one SV breakpoint overlaps a gnomAD exclusion region">

Implementation

1. Convert SV VCF to breakpoint BED format

breakpoint2BedSV \
  --vcf input.vcf \
  --output sv.breakpoints.bed

This step standardizes all SV representations (DEL/DUP/INV/BND/SVEND) into a unified breakpoint BED format.

---

2. Identify SVs overlapping gnomAD v4 exclusion regions

bedtools intersect \
  -a sv.breakpoints.bed \
  -b depth_blacklist.sorted.bed.gz \
  -wa | cut -f4 | sort -u > excluded_ids.txt

bedtools intersect \
  -a sv.breakpoints.bed \
  -b PESR.encode.peri_all.repeats.delly.hg38.blacklist.sorted.bed.gz \
  -wa | cut -f4 | sort -u >> excluded_ids.txt

tr "," "\n" < excluded_ids.txt | sort -u > excluded_ids.final.txt
rm excluded_ids.txt

---

3. Annotate original VCF with gnomAD_excl flag

awk -F'\t' '
BEGIN {
    OFS="\t"
    while ((getline line < "excluded_ids.final.txt") > 0)
        excl[line] = 1
}
{
    if ($0 ~ /^#/) {
        print
        next
    }

    id = $3

    if (id in excl) {
        if ($8 == "." || $8 == "") {
            $8 = "gnomAD_excl"
        } else {
            $8 = $8 ";gnomAD_excl"
        }
    }

    print
}
' input.vcf > input.gnomAD_excl.vcf

License

breakpoint2bedsv is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

breakpoint2bedsv is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

See the LICENSE file for the full license text.