hgvslib: A Python repository from Personalis, Inc.

Summary Information

Author: Jennifer Yen jennifer.lee.yen@gmail.com

Brief description

hgvslib is a Python package for parsing variant strings described according to recommendations by the Human Genome Variation Society (HGVS).

Purpose

This package was written to assess the accuracy and concordance of HGVS variant expressions between different annotation tools (Variant Effect Predictor, Variation Reporter, SnpEff) and the COSMIC and ClinVar databases. The results of these comparisons have been summarized in the manuscript 'A variant by any name: quantifying annotation discordance across tools and clinical databases': http://biorxiv.org/content/early/2016/05/19/054023.

Features

This package provides the following functionalities:

Assess whether two HGVS strings describe the same variant
Normalize an HGVS string to a reduced form
Parse attributes of a variant object (e.g. location, transcript, bases, etc)

Current limitations

Due to the complexity of HGVS syntax, this package currently does not support comparisons of HGVS descriptions that require querying the accession sequence. For example, although ‘NM_006015.4:c.492_494dupCGC’ and 'NM_006015.4:c.492_493insCGC’ describe the same variant, this package currently does not verify that CGC are duplicated bases according to the reference sequence.

The package was developed for the purpose of comparing to a reference ‘truth’ test set, from which deviations in positional justification are assumed to be non-HGVS compliant.

If there are HGVS expressions that are not currently supported by this tool, please make a comment or ping jennifer.lee.yen@gmail.com.

Future plans

Function to produce multiple HGVS aliases for the same variant from an expanded form
Cross-verify duplication and insertion sequences
Include HGVS validation test set (will be up soon)

Requirements

To install:

clone directory to your home directory git clone https://github.com/personalis/hgvslib.git
add cloned directory to your python path:

in ~./bashrc: PYTHONPATH=/path_to_hgvslib/hgvslib:$PYTHONPATH or
in python: sys.path.append('/path_to_hgvslib/hgvslib')

May need to install requirements: pip install -r requirements.txt

Unit Tests

Over 115 unit tests have been created to test this code.

To run:

python runtests.py -v

Example usage

from hgvslib.class_functions import compare_hgvs
import hgvslib.pHGVS as pHGVS

### Pandas usage

# Convert column of pHGVS syntax from singlet to triplet

df = pd.DataFrame({'mut_aa': ['p.T417_D419del2insY', 
                              'p.T46Hfs*35', 
                              'p.K355_K50del35insTPP', 
                              'p.N365_366insRDP', 
                              'p.R544X', 
                              'p.544fs36']})

# Convert column of singlet amino acid pHGVS to triplet form
df['mut_aa_triplet']        = df.mut_aa.apply(pHGVS.convert_phgvs_to_triplet, 1)

# Normalize pHGVS syntax to snpEFf minimal pHGVS syntax 
df['mut_aa_norm']           = df.mut_aa.apply(pHGVS.get_alias, 1)

# Get pHGVS variant consequence format
df['var_type']              = df.mut_aa.apply(pHGVS.get_var_type, 1)


# Get pHGVS variant consequence format
df['var_type']              = df.mut_aa.apply(pHGVS.get_var_type, 1)

# Check pHGVS nomenclature is correct (boolean)
df['is_correct']            = df.mut_aa.apply(pHGVS.is_proper_phgvs_format, 1)

### HGVS syntax comparisons

# The library can also perform basic parsing steps from an HGVS name.

from hgvslib.pHGVS import pHGVS

hgvs = pHGVS('p.Glu412fs*')
hgvs.alias
'p.Glu412fs'

hgvs.type
'frameshift'


from hgvslib.cHGVS import cHGVS

hgvs = cHGVS('NM_000352.3:c.215-10A>G')

hgvs.transcript
'NM_000352.3'

hgvs.name
'c.215-10A>G'

hgvs.type
'substitution'


# Basic transcript manipulation

hgvs1 = 'NM_004958.3'
t = Transcript(hgvs1)
t.name = 'NM_004958.3'
t.version = '3'
t.accession = 'NM_004958'




# The following check returns “yes”, indicating that this is an exact match.

hgvs1 = 'NM_000352.3:c.123delA'
hgvs2 = 'NM_000352.3:c.123delA'
compare_hgvs(hgvs1, hgvs2)
'yes'


# The following checks return “yes_m”, indicating that these are equivalent 
# but not exact matches ('yes modified').

hgvs1 = 'NP_005647.3:p.Q29fs'
hgvs2 = 'NP_005647.3:p.Q29Efs*10’
compare_hgvs(hgvs1, hgvs2)
'yes_m'

hgvs1 = 'p.Ser78Ser='
hgvs2 = 'p.='
compare_hgvs(hgvs1, hgvs2)
'yes_m'

hgvs1 = 'p.Glu78fs'
hgvs2 = 'p.Glu78GlyfsTer7'
compare_hgvs(hgvs1, hgvs2)
'yes_m'

hgvs1 = 'NM_004360.3:c.48+6_48+7delinsTT'
hgvs2 = 'NM_004360.3:c.48+6_48+7delCCinsTT'
compare_hgvs(hgvs1, hgvs2)
'yes_m'

# The following check returns “no”, indicating that these are different
# variants.

hgvs1 = 'NM_000352.3:c.123delA'
hgvs2 = 'NM_000352.3:c.125delA'
compare_hgvs(hgvs1, hgvs2)
'no'

hgvs1 = 'NP_005647.3:p.Q29fs'
hgvs2 = 'NP_005647.3:p.Q29Ter'
compare_hgvs(hgvs1, hgvs2)
'no'

Example script

The example script reads in a file to compare, for each variant, three different annotations with a reference HGVS string. The result of the comparison: an exact match 'yes', equivalent 'yes_m' or non-match 'no'.

Note: the example VCF file hgvs_test_cases.vcf has been normalized with vt-normalize https://github.com/atks/vt.

python example/example_check_variants.py --infile example/test_file.txt

personalis/hgvslib