Skip to content

Latest commit

 

History

History
105 lines (86 loc) · 5.91 KB

AlphaFold.md

File metadata and controls

105 lines (86 loc) · 5.91 KB

Alphafold 2.0 and 2.1 @ SAION

Disclaimer: this implementation was made for general use with the help of Jan Moren of SCS. I do take no responsibility for bugs or errors, but if you could report them, I will gladly try to fix on my extra time. Nonetheless, the AF2 predictions appear to be working. Feel free but not obligated to acknowledge me or at least the SCS if this work was helpful for your research.

In order to use AF2, you need SAION access and have basic knowledge of the terminal and slurm submission commands.
AF2 is part of the OIST Bioinfo User group.
AF2 was implemented using the docker image made by the University of Virginia, that here we gratefully acknowledge.

AF2 content

AF 2.0

  • 'alphafold_2.0.0.sif': singularity image created from the previously mentioned docker file
  • 'run_alphafold.sh': AF2 run script. Contains the default parameters (except input file and output folder). DO NOT MODIFY
  • 'alphafold_example_script.sh': AF2 sbatch example. Copy to your folder and modify (add input file and output folder paths).
  • 'get_plddt_attribfile.py': Chimera attribute file generator. Automatically generates an attribute file for plddt coloring in chimera.

AF 2.1

  • 'alphafold_2.1.1.sif': singularity image created from the previously mentioned docker file
  • 'run_alphafold.sh': AF2 run script. Contains the default parameters (except input file and output folder). DO NOT MODIFY
  • 'run_alphafold_multimer.sh': AF2 run script specialized in multimer run. Contains the default parameters (except input file and output folder). DO NOT MODIFY
  • 'alphafold_example_script.sh': AF2 sbatch example. Copy to your folder and modify (add input file and output folder paths).
  • 'get_plddt_attribfile.py': Chimera attribute file generator. Automatically generates an attribute file for plddt coloring in chimera.

MULTIMER RUN instructions TL;DR: provide a multi fasta file as input.

Example run

Assume to have a folder in /work with the input single protein sequence: example.fasta and a test output folder: test_out/.
Copy the example script:

cp /apps/unit/BioinfoUgrp/alphafold/2.1.1/bin/alphafold_example_script.sh .

Change the file name and modify the input and output folder paths, using your favourite text editor:

mv alphafold_example_script.sh AF2_testrun.sh
vim AF2_testrun.sh (change input file and output folder, save and exit)

OPTIONAL: CHECK IF THE RESOURCES YOU ARE REQUESTING VIA SLURM ARE ENOUGH (by default 4cpu cores, 100GB, 1 GPU)

Submit the job:

sbatch AF2_testrun.sh

The job is expected to take 2 to 12+ hours depending on sequence lenght.

Output

A description of the output can be found in AlphaFold 2 Github page.
However, for lazy people, ranked_0.pdb is the highest scoring model.

Example script

For convenience, we report here the example script content for version 2.1.1.

#!/bin/bash
#SBATCH --job-name=test_af
#SBATCH -c 4
#SBATCH --mem=100G
#SBATCH --partition=gpu
#SBATCH -t 4-0
#SBATCH --gres=gpu:1

module load bioinfo-ugrp-modules
module load alphafold/2.1.1

# Uncomment the monomer or multimer depending on the use case
# Monomer
# run_alphafold.sh --fasta_paths test.fa --output_dir test_output
# Multimer
# run_alphafold_multimer.sh --fasta_paths multifasta_test.fa --output_dir test_output

Attrib file

If you use Chimera to visualize the pdb file, you might be interested in per-residue plddt coloring. plddt is the self-evaluated confidence measure of the prediction of alphafold2. It goes from 0 (low confidence) to 100 (high confidence). A low value indicates either badly predicted regions or disordered regions.
To automatically generate a chimera attribute file for coloring by plddt, you can use the python script I made:

module load python/3.7.3
python /apps/unit/BioinfoUgrp/alphafold/2.0.0/bin/get_plddt_attribfile.py <result_model_X.pkl> <outfile.txt>
The text file generated can be used in Chimera (Tools > Structure Analysis > Define Attribute) after opening the same pdb model.

Appendix: parameters

This is a list of all the default parameters in the run_alphafold script. If you want to modify them, do not change the default script. Instead, provide them to the example run script just like the input fasta and output folder.
Default parameters (v2.0):

MAX_TEMPLATE_DATE=2030-09-09
MODEL_NAMES=model_1,model_2,model_3,model_4,model_5
DATA_DIR=/apps/unit/BioinfoUgrp/alphafold/database/
UNIREF90_PATH=$DATA_DIR/uniref90/uniref90.fasta
MGNIFY_PATH=$DATA_DIR/mgnify/mgy_clusters.fa
UNICLUST30_PATH=$DATA_DIR/uniclust30/uniclust30_2018_08/uniclust30_2018_08
BFD_PATH=$DATA_DIR/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt
PDB70_PATH=$DATA_DIR/pdb70/pdb70
TEMPLATE_PATH=$DATA_DIR/pdb_mmcif/mmcif_files/
OBSOLETE_PATH=$DATA_DIR/pdb_mmcif/obsolete.dat

Default parameters (v2.1):

MAX_TEMPLATE_DATE=2030-09-09
MODEL_PRESET=multimer
DATA_DIR=/apps/unit/BioinfoUgrp/alphafold/database/
UNIREF90_PATH=$DATA_DIR/uniref90/uniref90.fasta
MGNIFY_PATH=$DATA_DIR/mgnify/mgy_clusters.fa
UNICLUST30_PATH=$DATA_DIR/uniclust30/uniclust30_2018_08/uniclust30_2018_08
BFD_PATH=$DATA_DIR/bfd/bfd_metaclust_clu_complete_id30_c90_final_seq.sorted_opt
PDB70_PATH=$DATA_DIR/pdb70/pdb70
TEMPLATE_PATH=$DATA_DIR/pdb_mmcif/mmcif_files/
OBSOLETE_PATH=$DATA_DIR/pdb_mmcif/obsolete.dat
PDB_SEQRES_PATH=$DATA_DIR/pdb_seqres/pdb_seqres.txt
UNIPROT_PATH=$DATA_DIR/uniprot/

Appendix: contact

Stefano Pascarelli, PhD, Laurino unit