AlphaFold is a deep learning tool for predicting a protein’s 3D structure from its amino acid sequence. It was developed by DeepMind and utilizes GPU compute. For more information, see the tool's homepage and GitHub repo.

Function Call

tc.alphafold(
    inputs,
    output_path=None,
    model_preset=None,
    max_template_date=None,
    use_reduced_dbs=False,
    is_prokaryote_list=None,
    is_async=False,
)

Function Arguments

See the Notes section below for more details.

Argument	Use in place of:	Description
inputs	--fasta-paths	Path to one or more files to use as input. The files can be a local or remote, see Using Files.
output_path	--output_dir	(optional) Path (directory) to where the output files will be downloaded. If omitted, skips download. The files can be a local or remote, see Using Files.
model_preset	--model_preset	(optional) Specific AlphaFold model to use. Options are [monomer, monomer_casp14, monomer_ptm, multimer]. Defaults to monomer.
max_template_date	--max_template_date	(optional) String of date in YYYY-MM-DD format. Restricts protein structure prediction to those in the database before this date. Defaults to today's date.
use_reduced_dbs	--db_preset=reduced_dbs	(optional) Whether to use a smaller version of the BFD database. If true, reduces run time at the cost of result quality.
is_prokaryote_list	--is_prokaryote_list	(optional) List of booleans that determines whether all input sequences in the given FASTA file are prokaryotic. Expects the string normally used input into AlphaFold (e.g. "true,true" if there are two prokaryote inputs).
is_async		Whether to run a job asynchronously. See Async Runs for more.

Tool Versions

Toolchest currently supports version 2.1.2 of AlphaFold.

Database

Toolchest's implementation of AlphaFold uses AlphaFold's required genetic sequence databases. For a complete list of databases used, see the tool's GitHub page.

Supported Additional Arguments

Toolchest supports the following arguments for AlphaFold:

• --db_preset
• --is_prokaryote_list
• --max_template_date
• --model_preset

However, these should be specified via specific argument values in the function call, rather than through a generic tool_args argument (like other Toolchest tools). See Function Arguments for more details.