cyvcf2 0.30.26

fast vcf parsing with cython + htslib

cyvcf2

Note: cyvcf2 versions < 0.20.0 require htslib < 1.10. cyvcf2 versions >= 0.20.0 require htslib >= 1.10

The latest documentation for cyvcf2 can be found here:

If you use cyvcf2, please cite the paper

Fast python (2 and 3) parsing of VCF and BCF including region-queries.

cyvcf2 is a cython wrapper around htslib built for fast parsing of Variant Call Format (VCF) files.

Attributes like variant.gt_ref_depths work for diploid samples and return a numpy array directly so they are immediately ready for downstream use. note that the array is backed by the underlying C data, so, once variant goes out of scope. The array will contain nonsense. To persist a copy, use: cpy = np.array(variant.gt_ref_depths) instead of just arr = variant.gt_ref_depths.

Example

The example below shows much of the use of cyvcf2.

from cyvcf2 import VCF

for variant in VCF('some.vcf.gz'): # or VCF('some.bcf')
    variant.REF, variant.ALT # e.g. REF='A', ALT=['C', 'T']

    variant.CHROM, variant.start, variant.end, variant.ID, \
                variant.FILTER, variant.QUAL

    # numpy arrays of specific things we pull from the sample fields.
    # gt_types is array of 0,1,2,3==HOM_REF, HET, UNKNOWN, HOM_ALT
    variant.gt_types, variant.gt_ref_depths, variant.gt_alt_depths # numpy arrays
    variant.gt_phases, variant.gt_quals, variant.gt_bases # numpy array

    ## INFO Field.
    ## extract from the info field by it's name:
    variant.INFO.get('DP') # int
    variant.INFO.get('FS') # float
    variant.INFO.get('AC') # float

    # convert back to a string.
    str(variant)


    ## sample info...

    # Get a numpy array of the depth per sample:
    dp = variant.format('DP')
    # or of any other format field:
    sb = variant.format('SB')
    assert sb.shape == (n_samples, 4) # 4-values per

# to do a region-query:

vcf = VCF('some.vcf.gz')
for v in vcf('11:435345-556565'):
    if v.INFO["AF"] > 0.1: continue
    print(str(v))

Installation

pip (assuming you have htslib < 1.10 installed)

pip install cyvcf2

github (building htslib and cyvcf2 from source)

git clone --recursive https://github.com/brentp/cyvcf2
pip install -r requirements.txt
CYVCF2_HTSLIB_MODE=BUILTIN CYTHONIZE=1 python setup.py install
# or to use a system htslib.so
CYVCF2_HTSLIB_MODE=EXTERNAL python setup.py install

On OSX, using brew, you may have to set the following as indicated by the brew install:

For compilers to find openssl you may need to set:
  export LDFLAGS="-L/usr/local/opt/openssl/lib"
  export CPPFLAGS="-I/usr/local/opt/openssl/include"

For pkg-config to find openssl you may need to set:
  export PKG_CONFIG_PATH="/usr/local/opt/openssl/lib/pkgconfig"

Testing

Install pytest, then tests can be run with:

pytest

CLI

Run with cyvcf2 path_to_vcf

$ cyvcf2 --help
Usage: cyvcf2 [OPTIONS] <vcf_file> or -

  fast vcf parsing with cython + htslib

Options:
  -c, --chrom TEXT                Specify what chromosome to include.
  -s, --start INTEGER             Specify the start of region.
  -e, --end INTEGER               Specify the end of the region.
  --include TEXT                  Specify what info field to include.
  --exclude TEXT                  Specify what info field to exclude.
  --loglevel [DEBUG|INFO|WARNING|ERROR|CRITICAL]
                                  Set the level of log output.  [default:
                                  INFO]
  --silent                        Skip printing of vcf.
  --help                          Show this message and exit.

See Also

Pysam also has a cython wrapper to htslib and one block of code here is taken directly from that library. But, the optimizations that we want for gemini are very specific so we have chosen to create a separate project.

Performance

For the performance comparison in the paper, we used thousand genomes chromosome 22 With the full comparison runner here.