Metadata-Version: 2.1
Name: vquest
Version: 0.0.4
Summary: Automate IMGT V-QUEST usage on imgt.org
Home-page: https://github.com/ressy/vquest
Author: Jesse Connell
Author-email: jesse@ressy.us
License: UNKNOWN
Description: # Automate IMGT V-QUEST usage on imgt.org
        
        [IMGT](http://imgt.org)'s [V-QUEST](http://www.imgt.org/IMGT_vquest/analysis)
        is only available via a web interface.  This Python package automates V-QUEST
        usage by submitting request data like the web form does.  Curently only the
        "Download AIRR formatted results" option is supported.
        
        Example command-line usage, with rhesus sequences in seqs.fasta:
        
            pip install . # or skip this and use "python -m vquest" below
            vquest --species rhesus-monkey --receptorOrLocusType IG --fileSequences seqs.fasta
        
        The output is saved to `Parameters.txt` and `vquest_airr.tsv` (the files
        V-QUEST provides in a zip archive) in the working directory.
        
        Example Python usage:
        
            >>> from vquest import *
            >>> config = layer_configs(DEFAULTS, {"species": "rhesus-monkey", "receptorOrLocusType": "IG", "fileSequences": "seqs.fasta"})
            >>> result = vquest(config)
            >>> result.keys()
            dict_keys(['Parameters.txt', 'vquest_airr.tsv'])
        
        Here the output is a dictionary of filenames to contents.
        
        The only required options are species, receptorOrLocusType, and either
        fileSequences or sequences (to provide sequences directly as text).  Options
        can be given via command-line arguemnts or one or more YAML configuration
        files.  See [data/defaults.yml](data/defaults.yml) and `./vquest.py --help` for
        details.
        
        The web form will only accept 50 sequences at a time, so the sequences given
        here are grouped into chunks of 50, submitted, and the results combined.  A
        delay (default 1 second) is used between submissions to avoid being impolite to
        the server.
        
         * V-QUEST: <http://www.imgt.org/IMGT_vquest/analysis>
         * V-QUEST docs: <http://www.imgt.org/IMGT_vquest/user_guide#intro>
         * A different approach, using [Selenium](https://www.selenium.dev/) to automate V-QUEST usage with a browser: <https://github.com/AndrewZoldy/IMGT_VQUEST_BOT>
        
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: GNU Affero General Public License v3
Classifier: Operating System :: OS Independent
Requires-Python: >=3.6
Description-Content-Type: text/markdown
