Awesome

Foldseek

Foldseek enables fast and sensitive comparisons of large protein structure sets.

Publications

van Kempen M, Kim S, Tumescheit C, Mirdita M, Lee J, Gilchrist CLM, Söding J, and Steinegger M. Fast and accurate protein structure search with Foldseek. Nature Biotechnology, doi:10.1038/s41587-023-01773-0 (2023)

Barrio-Hernandez I, Yeo J, Jänes J, Mirdita M, Gilchrist CLM, Wein T, Varadi M, Velankar S, Beltrao P and Steinegger M. Clustering predicted structures at the scale of the known protein universe. Nature, doi:10.1038/s41586-023-06510-w (2023)

Kim W, Mirdita M, Levy Karin E, Gilchrist CLM, Schweke H, Söding J, Levy E, and Steinegger M. Rapid and Sensitive Protein Complex Alignment with Foldseek-Multimer. bioRxiv, doi:10.1101/2024.04.14.589414 (2024)

Table of Contents

• Foldseek
  • Publications
• Table of Contents
  • Webserver
  • Installation
  • Memory requirements
  • Tutorial Video
  • Documentation
  • Quick start
    • Search
      • Output Search
        • Tab-separated
        • Superpositioned Cα only PDB files
        • Interactive HTML
      • Important search parameters
      • Alignment Mode
      • Structure search from FASTA input
    • Databases
      • Create custom databases and indexes
    • Cluster
      • Output Cluster
        • Tab-separated cluster
        • Representative fasta
        • All member fasta
      • Important cluster parameters
    • Multimersearch
      • Using Multimersearch
      • Multimer Search Output
        • Tab-separated-complex
        • Complex Report
    • Multimercluster
      • Output MultimerCluster
        • Tab-separated multimercluster
        • Representative multimer fasta
        • Filtered search result
      • Important multimer cluster parameters
  • Main Modules
  • Examples
    • Rescore aligments using TMscore
    • Query centered multiple sequence alignment

Webserver

Search your protein structures against the AlphaFoldDB and PDB in seconds using the Foldseek webserver (code): search.foldseek.com 🚀

Installation

# Linux AVX2 build (check using: cat /proc/cpuinfo | grep avx2)
wget https://mmseqs.com/foldseek/foldseek-linux-avx2.tar.gz; tar xvzf foldseek-linux-avx2.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

# Linux SSE2 build (check using: cat /proc/cpuinfo | grep sse2)
wget https://mmseqs.com/foldseek/foldseek-linux-sse2.tar.gz; tar xvzf foldseek-linux-sse2.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

# Linux ARM64 build
wget https://mmseqs.com/foldseek/foldseek-linux-arm64.tar.gz; tar xvzf foldseek-linux-arm64.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

# MacOS
wget https://mmseqs.com/foldseek/foldseek-osx-universal.tar.gz; tar xvzf foldseek-osx-universal.tar.gz; export PATH=$(pwd)/foldseek/bin/:$PATH

# Conda installer (Linux and macOS)
conda install -c conda-forge -c bioconda foldseek

Other precompiled binaries for ARM64 amd SSE2 are available at https://mmseqs.com/foldseek.

Memory requirements

For optimal software performance, consider three options based on your RAM and search requirements:

1. With Cα info (default). Use this formula to calculate RAM - (6 bytes Cα + 1 3Di byte + 1 AA byte) * (database residues). The 54M AFDB50 entries require 151GB.

2. Without Cα info. By disabling --sort-by-structure-bits 0, RAM requirement reduces to 35GB. However, this alters hit rankings and final scores but not E-values. Structure bits are mostly relevant for hit ranking for E-value > 10^-1.

3. Single query searches. Use the --prefilter-mode 1, which isn't memory-limited and computes all ungapped alignments. This option optimally utilizes foldseek's multithreading capabilities for single queries.

Tutorial Video

We presented a Foldseek tutorial at the SBGrid where we demonstrated Foldseek's webserver and command line interface. Check it out here.

<a href="https://www.youtube.com/watch?v=k5Rbi22TtOA"><img src="https://img.shields.io/youtube/views/k5Rbi22TtOA?style=social"></a>.

Documentation

Many of Foldseek's modules (subprograms) rely on MMseqs2. For more information about these modules, refer to the MMseqs2 wiki. For documentation specific to Foldseek, checkout the Foldseek wiki here.

Quick start

Search

The easy-search module allows to query one or more single-chain protein structures, formatted in PDB/mmCIF format (flat or gzipped), against a target database, folder or individual single-chain protein structures (for multi-chain proteins see complexsearch). The default alignment information output is a tab-separated file but Foldseek also supports Superposed Cα PDBs and HTML.

foldseek easy-search example/d1asha_ example/ aln tmpFolder

Output Search

Tab-separated

The default output fields are: query,target,fident,alnlen,mismatch,gapopen,qstart,qend,tstart,tend,evalue,bits but they can be customized with the --format-output option e.g., --format-output "query,target,qaln,taln" returns the query and target accessions and the pairwise alignments in tab-separated format. You can choose many different output columns.

Check out the MMseqs2 documentation for additional output format codes.

Superpositioned Cα only PDB files

Foldseek's --format-mode 5 generates PDB files with all target Cα atoms superimposed onto the query structure based on the aligned coordinates. For each pairwise alignment it will write its own PDB file, so be careful when using this options for large searches.

Interactive HTML

Locally run Foldseek can generate an HTML search result, similar to the one produced by the webserver by specifying --format-mode 3

foldseek easy-search example/d1asha_ example/ result.html tmp --format-mode 3

Important search parameters

Alignment Mode

By default, Foldseek uses its local 3Di+AA structural alignment but it also supports realigning hits using the global TMalign as well as rescoring alignments using TMscore.

foldseek easy-search example/d1asha_ example/ aln tmp --alignment-type 1

If alignment type is set to tmalign (--alignment-type 1), the results will be sorted by the TMscore normalized by query length. The TMscore is used for reporting two fields: the e-value=(qTMscore+tTMscore)/2 and the score=(qTMscore*100). All output fields (e.g., pident, fident, and alnlen) are calculated based on the TMalign alignment.

Structure search from FASTA input

Search by predicting 3Di directly from amino acid sequences without the need for existing protein structures. This feature uses the ProstT5 protein language model and runs by default on CPU and is about 400-4000x compared to predicted structures by ColabFold.

foldseek databases ProstT5 weights tmp
foldseek databases PDB pdb tmp
foldseek easy-search QUERY.fasta pdb result.m8 tmp --prostt5-model weights

Or create your a structural database from a fasta files.

foldseek createdb db.fasta db --prostt5-model weights

Faster inference using GPU/CUDA is also supported. Compile from source with cmake -DCMAKE_BUILD_TYPE=Release -DENABLE_CUDA=1 -DCUDAToolkit_ROOT=Path-To-Cuda-Toolkit and call with createdb/easy-search --prostt5-model weights --gpu 1.

Databases

The databases command downloads pre-generated databases like PDB or AlphaFoldDB.

# pdb  
foldseek databases PDB pdb tmp 
# alphafold db
foldseek databases Alphafold/Proteome afdb tmp 

We currently support the following databases:

  Name                   	Type     	Taxonomy	Url
- Alphafold/UniProt   	Aminoacid	     yes	https://alphafold.ebi.ac.uk/
- Alphafold/UniProt50 	Aminoacid	     yes	https://alphafold.ebi.ac.uk/
- Alphafold/Proteome  	Aminoacid	     yes	https://alphafold.ebi.ac.uk/
- Alphafold/Swiss-Prot	Aminoacid	     yes	https://alphafold.ebi.ac.uk/
- ESMAtlas30          	Aminoacid	       -	https://esmatlas.com
- PDB                 	Aminoacid	     yes	https://www.rcsb.org

Create custom databases and indexes

The target database can be pre-processed by createdb. This is useful when searching multiple times against the same set of target structures.

foldseek createdb example/ targetDB
foldseek createindex targetDB tmp  #OPTIONAL generates and stores the index on disk
foldseek easy-search example/d1asha_ targetDB aln.m8 tmpFolder

Cluster

The easy-cluster algorithm is designed for structural clustering by assigning structures to a representative protein structure using structural alignment. It accepts input in either PDB or mmCIF format, with support for both flat and gzipped files. By default, easy-cluster generates three output files with the following prefixes: (1) _clu.tsv, (2) _repseq.fasta, and (3) _allseq.fasta. The first file (1) is a tab-separated file describing the mapping from representative to member, while the second file (2) contains only representative sequences, and the third file (3) includes all cluster member sequences.

foldseek easy-cluster example/ res tmp -c 0.9 

Output Cluster

Tab-separated cluster

The provided format represents protein structure clustering in a tab-separated, two-column layout (representative and member). Each line denotes a cluster-representative and cluster-member relationship, signifying that the member shares significant structural similarity with the representative, and thus belongs to the same cluster.

Q0KJ32	Q0KJ32
Q0KJ32	C0W539
Q0KJ32	D6KVP9
E3HQM9	E3HQM9
E3HQM9	F0YHT8

Representative fasta

The _repseq.fasta contains all representative protein sequences of the clustering.

>Q0KJ32
MAGA....R
>E3HQM9
MCAT...Q

All member fasta

In the _allseq.fasta file all sequences of the cluster are present. A new cluster is marked by two identical name lines of the representative sequence, where the first line stands for the cluster and the second is the name line of the first cluster sequence. It is followed by the fasta formatted sequences of all its members.

>Q0KJ32	
>Q0KJ32
MAGA....R
>C0W539
MVGA....R
>D6KVP9
MVGA....R
>D1Y890
MVGV....R
>E3HQM9	
>E3HQM9
MCAT...Q
>Q223C0
MCAR...Q

Important cluster parameters

Multimersearch

The easy-multimersearch module is designed for querying one or more protein complex (multi-chain) structures (supported input formats: PDB/mmCIF, flat or gzipped) against a target database of protein complex structures. It reports the similarity metrices between the complexes (e.g., the TMscore).

Using Multimersearch

The examples below use files that can be found in the example directory, which is part of the Foldseek repo, if you clone it. If you use the precompiled version of the software, you can download the files directly: 1tim.pdb.gz and 8tim.pdb.gz.

For a pairwise alignment of complexes using easy-multimersearch, run the following command:

foldseek easy-multimersearch example/1tim.pdb.gz example/8tim.pdb.gz result tmpFolder

Foldseek easy-multimersearch can also be used for searching one or more query complexes against a target database:

foldseek databases PDB pdb tmp 
foldseek easy-multimersearch example/1tim.pdb.gz pdb result tmpFolder

Multimer Search Output

Tab-separated-complex

By default, easy-multimersearch reports the output alignment in a tab-separated file. The default output fields are: query,target,fident,alnlen,mismatch,gapopen,qstart,qend,tstart,tend,evalue,bits,complexassignid but they can be customized with the --format-output option e.g., --format-output "query,target,complexqtmscore,complexttmscore,complexassignid" alters the output to show specific scores and identifiers.

Example Output:

1tim.pdb.gz_A   8tim.pdb.gz_A   0.967   247 8   0   1   247 1   247 5.412E-43   1527    0
1tim.pdb.gz_B   8tim.pdb.gz_B   0.967   247 8   0   1   247 1   247 1.050E-43   1551    0

Complex Report

easy-multimersearch also generates a report (prefixed _report), which provides a summary of the inter-complex chain matching, including identifiers, chains, TMscores, rotation matrices, translation vectors, and assignment IDs. The report includes the following fields:

Example Output:

1tim.pdb.gz 8tim.pdb.gz A,B A,B 0.98941 0.98941 0.999983,0.000332,0.005813,-0.000373,0.999976,0.006884,-0.005811,-0.006886,0.999959 0.298992,0.060047,0.565875  0

Multimercluster

The easy-multimercluster module is designed for multimer-level structural clustering(supported input formats: PDB/mmCIF, flat or gzipped). By default, easy-multimercluster generates three output files with the following prefixes: (1) _cluster.tsv, (2) _rep_seq.fasta and (3) _cluster_report. The first file (1) is a tab-separated file describing the mapping from representative multimer to member, while the second file (2) contains only representative sequences. The third file (3) is also a tab-separated file describing filtered alignments.

Make sure chain names in PDB/mmcIF files does not contain underscores(_).

foldseek easy-multimercluster example/ clu tmp --multimer-tm-threshold 0.65 --chain-tm-threshold 0.5 --interface-lddt-threshold 0.65

Output MultimerCluster

Tab-separated multimercluster

5o002	   5o002
194l2	   194l2
194l2	   193l2
10mh121	 10mh121
10mh121	 10mh114
10mh121	 10mh119

Representative multimer fasta

#5o002
>5o002_A
SHGK...R
>5o002_B
SHGK...R
#194l2
>194l2_A0
KVFG...L
>194l2_A6
KVFG...L
#10mh121
...

Filtered search result

The _cluster_report contains qcoverage, tcoverage, multimer qTm, multimer tTm, interface lddt, ustring, tstring of alignments after filtering and before clustering.

5o0f2	5o0f2	1.000	1.000	1.000	1.000	1.000	1.000,0.000,0.000,0.000,1.000,0.000,0.000,0.000,1.000	0.000,0.000,0.000
5o0f2	5o0d2	1.000	1.000	0.999	0.992	1.000	0.999,0.000,-0.000,-0.000,0.999,-0.000,0.000,0.000,0.999	-0.004,-0.001,0.084
5o0f2	5o082	1.000	0.990	0.978	0.962	0.921	0.999,-0.025,-0.002,0.025,0.999,-0.001,0.002,0.001,0.999	-0.039,0.000,-0.253

The query and target coverages here represent the sum of the coverages of all aligned chains, divided by the total query and target multimer length respectively.

Important multimer cluster parameters

Main Modules

• easy-search fast protein structure search
• easy-cluster fast protein structure clustering
• easy-multimersearch fast protein multimer-level structure search
• easy-multimercluster fast protein multimer-level structure clustering
• createdb create a database from protein structures (PDB,mmCIF, mmJSON)
• databases download pre-assembled databases

Examples

Rescore aligments using TMscore

The easiest way to get the alignment TMscore normalized by min(alnLen,qLen,targetLen) as well as a rotation matrix is through the following command:

foldseek easy-search example/ example/ aln tmp --format-output query,target,alntmscore,u,t

Alternatively, it is possible to compute TMscores for the kind of alignment output (e.g., 3Di+AA) using the following commands:

foldseek createdb example/ targetDB
foldseek createdb example/ queryDB
foldseek search queryDB targetDB aln tmpFolder -a
foldseek aln2tmscore queryDB targetDB aln aln_tmscore
foldseek createtsv queryDB targetDB aln_tmscore aln_tmscore.tsv

Output format aln_tmscore.tsv: query and target identifiers, TMscore, translation(3) and rotation vector=(3x3)

Query centered multiple sequence alignment

Foldseek can output multiple sequence alignments in a3m format using the following commands. To convert a3m to FASTA format, the following script can be used reformat.pl (reformat.pl in.a3m out.fas).

foldseek createdb example/ targetDB
foldseek createdb example/ queryDB
foldseek search queryDB targetDB aln tmpFolder -a
foldseek result2msa queryDB targetDB aln msa --msa-format-mode 6
foldseek unpackdb msa msa_output --unpack-suffix a3m --unpack-name-mode 0