References:

• M. van Kempen, S. Kim, C. Tumescheit, M. Mirdita, J. Söding, and M. Steinegger. Foldseek: fast and accurate protein structure search. BioRxiv, 2022 doi:10.1101/2022.02.07.479398

• foldseek onGitHub
• foldseek on the web at search.foldseek.com

• Module Name: foldseek (see the modules page for more information)
• foldseek is multithreaded and not slurm aware. Please specify threads to match the number of allocated CPUs to avoid overloaded jobs
• Example files in $FOLDSEEK_TEST_DATA
• Reference data in $FOLDSEEK_DB/

2024-12-12: Version 9 installed

Version 9 introduced a GPU accellerated structure search mode directly from fasta files using the ProstT5 protein language model. See here. Model weights are found in $FOLDSEEK_DB/ProstT5

Allocate an interactive session for a short tutorial:

[user@biowulf]$ sinteractive --mem=24g --cpus-per-task=12 --gres=lscratch:20
salloc.exe: Pending job allocation 46116226
salloc.exe: job 46116226 queued and waiting for resources
salloc.exe: job 46116226 has been allocated resources
salloc.exe: Granted job allocation 46116226
salloc.exe: Waiting for resource configuration
salloc.exe: Nodes cn3144 are ready for job

[user@cn3144]$ cd /lscratch/$SLURM_JOB_ID
[user@cn3144]$ module load foldseek/5-53465f0

Create a database from PDB or mmCif files to speed up repeated searches

[user@cn3144]$ foldseek createdb --threads $SLURM_CPUS_PER_TASK $FOLDSEEK_TEST_DATA testDB
MMseqs Version:         3c64211f59830702e2a369d6e6f0d8a7492c27fa
Chain name mode         0
Write lookup file       1
Threads                 12
Verbosity               3

Output file: testDB
[=================================================================] 100.00% 26 0s 57ms
Time for merging to testDB_ss: 0h 0m 0s 0ms
Time for merging to testDB_h: 0h 0m 0s 0ms
Time for merging to testDB_ca: 0h 0m 0s 0ms
Time for merging to testDB: 0h 0m 0s 0ms
Ignore 0 out of 26.
Too short: 0, incorrect  0.
Time for processing: 0h 0m 0s 94ms

[user@cn3144]$ foldseek easy-search --threads $SLURM_CPUS_PER_TASK \
    $FOLDSEEK_TEST_DATA/d1asha_ testDB aln.m8 /lscratch/$SLURM_JOB_ID/tmp
...
[user@cn3144]$ head aln.m8
d1asha_ d1asha_ 1.000   147     0       0       1       147     1       147     2.859E-22       1061
d1asha_ d1x9fd_ 0.173   143     111     0       3       145     5       139     2.716E-05       265
d1asha_ d2w72b_ 0.182   145     112     0       1       145     4       141     4.463E-04       208
d1asha_ d1itha_ 0.131   145     119     0       1       145     3       140     6.611E-04       200
d1asha_ d1mbaa_ 0.118   143     120     0       4       146     6       142     6.611E-04       200

There are also pre-build foldseek databases for AlphafoldDB and PDB in $FOLDSEEK_DB.

[user@cn3144]$ foldseek easy-search --threads $SLURM_CPUS_PER_TASK \
    --format-mode 4 \
    --format-output query,target,taxid,taxname,fident,alnlen,mismatch,qstart,qend,tstart,tend,evalue \
    $FOLDSEEK_TEST_DATA/d1asha_ $FOLDSEEK_DB/Alphafold-SwissProt aln_afdb.tsv /lscratch/$SLURM_JOB_ID/tmp
[user@cn3144]$

[user@cn3144]$ exit
salloc.exe: Relinquishing job allocation 46116226
[user@biowulf]$

In the following example individual steps are carried out separately and hits are re-aligned with TM-align. As an alternative --alignment-type 1 for easy-search uses TM-align and write the TMscore into the evalue field. TM-score is in the range of (0,1]. 1 indicates a perfect match between two structures. Scores below 0.17 correspond to randomly chosen unrelated proteins whereas structures with a score higher than 0.5 assume generally the same fold. The .tsv formatted outputs below can be merged to combine results.

[user@cn3144]$ foldseek createdb --threads $SLURM_CPUS_PER_TASK $FOLDSEEK_TEST_DATA queryDB
[user@cn3144]$ foldseek search -s 9.5 -a --threads $SLURM_CPUS_PER_TASK queryDB $FOLDSEEK_DB/PDB alnDB /lscratch/$SLURM_JOB_ID/tmp
[user@cn3144]$ foldseek convertalis --threads $SLURM_CPUS_PER_TASK queryDB $FOLDSEEK_DB/PDB alnDB aln.tsv
[user@cn3144]$ foldseek aln2tmscore --threads $SLURM_CPUS_PER_TASK queryDB $FOLDSEEK_DB/PDB alnDB alntmDB
[user@cn3144]$ foldseek createtsv --threads $SLURM_CPUS_PER_TASK queryDB $FOLDSEEK_DB/PDB alntmDB aln_tm.tsv
[user@cn3144]$ exit
salloc.exe: Relinquishing job allocation 46116226
[user@biowulf]$

Create a batch input file (e.g. foldseek.sh), which uses the input file 'foldseek.in'. For example:

#!/bin/bash
module load foldseek/1-2dd3b2f
foldseek easy-search --threads $SLURM_CPUS_PER_TASK \
    --format-mode 4 \
    --format-output query,target,taxid,taxname,fident,alnlen,mismatch,qstart,qend,tstart,tend,evalue \
    $FOLDSEEK_TEST_DATA/d1asha_ $FOLDSEEK_DB/Alphafold-SwissProt aln_afdb.tsv /lscratch/$SLURM_JOB_ID/tmp

Submit this job using the Slurm sbatch command.

sbatch --cpus-per-task=6 --mem=6G foldseek.sh