ghfc-utils 0.0.1

Various tools and scripts used in the GHFC lab

ghfc-utils

set of small tools designed to help automatize simple task locally or on Pasteur's cluster.

ghfc-reannotate for the postprocessing of slivar files including filtering and geneset reannotation.

Installation

pip install git+https://{token_username}:{generated_token}@gitlab.pasteur.fr/ghfc/ghfc-utils.git

when on maestro:

module load Python/3.9.18
pip install --user git+https://maestro_tok_FC:[e-mail]/ghfc/ghfc-utils.git

slivar reannotator

A tool to filter and reannotate slivar files according to various parameters and genesets. The goal is to produce a more generic kind of slivar files and to use this for the user to run their own filtering.

usage: ghfc-reannotate [-h] configuration slivar output

positional arguments:
  configuration         config file
  slivar                slivar file to reannotate
  output                annotated slivar file

optional arguments:
  -h, --help            show this help message and exit
  --chunksize CHUNKSIZE
                        size of the chunks read from the input (default 100000)

• This tools read the slivar file before decomposing the impacts by transcripts.
• It then filters all line using, in this order following the config file parameters:
  1. the geneset (based on the ENSG, mind the GRCh37/GRCh38 differences in ENSG)
  2. the impact / impact-categories
  3. if missense are kept, filtering them on their impact (using scores such as the mpc or the cadd)
  4. the gnomad frequency
• the variant/transcript are then sorted according to criteria given by the user in the config file from the most important to the least important
• for each sample, variant and gene (ENSG) the first transcript (most important given the config criteria) is kept.

yaml config file

This section is listing the accepted options in the config file, but example files are provided.

• (optional) geneset-file: path to the file containing the list of ENSG for out geneset of interest
• ordering-priority: the list (ordered) of criteria to use to rank the importance of the transcripts to output
• impact-categories-filter: the impact categories to keep during the filtering. The categories are defined in this package in impacts.yaml.
• impact-filter: the impacts to keep during the filtering process. the name of the impacts are visible in impacts.yaml.
• (optional) missense-filter: this section will define how the missense variant are further filtered
  • first a subcategory is used for the score used, e.g. mpc, cadd
  • for each subcategory, 3 fields are expected:
    • field: the name of the slivar column containing the value
    • min: the minimal value to keep (included)
    • max: the maximal value to keep (excluded)
  • in addition to the subcategories, a condition field is expected to specify how the subcategories are used. Possible values are:
    • cadd_if_no_mpc: use the mpc and when not available (-1) uses the cadd
    • cadd_and_mpc
    • cadd_or_mpc
    • mpc_only
    • cadd_only
• (optional) gnomad-filter: to filter further on an included gnomAD column. 3 fields are expected here:
  • field: the name of the slivar column containing the gnomad value to filter on
  • min: the minimal value to keep (included)
  • max: the maximal value to keep (included)
• (optional) pext-filter: to annotate each transcript and filter them using a pext file:
  • file: path to the pext file to use
  • field: nme of the outputed column
  • min: the minimal value to keep (included). can put -1 to annotate and not filter on it.

Finally, some more global slivar parameters that are not likely to change a lot:

• slivar-field-name: the name of the slivar column that contains the list of all vep impacts per transcripts
• slivar-field-decomposed: the list of each field when they are decomposed. some of those fields are expected with the following names:
  • impact
  • ENSG
  • canonical: the vep columns containing "YES" for the canonical transcripts
  • loftee: the loftee "LoF" column

pext file

The pext is a bed file with the following columns (order important, there must be some header):

chr	start	end	max_brain	ensg	symbol

Need to have the genome version to match the data (GRCh37/38 and using the chr or not in the chromosome names)

TODO

• offer to prefix geneset columns?
• offer to keep some of the original columns (impact/transcript)
• possibility to run on stdin / stdout?
• refining DP and AB
• needs for automated submission on the cluster? (means user has permission to use it)
• possibility to automate splitting in chunks and merging back?

slivar de novo ML

Moving the machine learning validator for de novo variants to this tool.