AlphaFold

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

toolchest.alphafold(
    inputs,
    output_path=None,
    model_preset=None,
    max_template_date=None,
    use_reduced_dbs=False,
    is_prokaryote_list=None,
    is_async=False,
)
toolchest$alphafold(
    inputs,
    output_path = NULL,
    model_preset = NULL,
    max_template_date = NULL,
    use_reduced_dbs = FALSE,
    is_prokaryote_list = NULL,
    is_async = FALSE
)

Function Arguments

See the Notes section below for more details.

Argument

Use in place of:

Description

inputs

--fasta-paths

Path or list of paths to file(s) that will be passed in as input. A path can be a local filepath or an AWS S3 URI.

output_path

--output_dir

(optional) Path to a local directory where the output files will be downloaded.
If omitted, skips download. The outputs can be downloaded manually.

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.

Return Value

This function call returns a Toolchest output object, which contains the run ID and locations of downloaded output files. See Output Objects for more details.

Notes

Amazon AWS S3 inputs

Publicly available files stored on AWS's S3 service can be passed in as inputs, using the file's S3 URI.

Async runs

Set the is_async parameter to true if you would like to run a AlphaFold job asynchronously. See Async Runs for more information.

Tool Versions

Toolchest currently supports version 2.1.2 of AlphaFold. Every request to run AlphaFold with Toolchest will default to this version.

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.


Did this page help you?