ColabFold

+ 2022/01/03: The MSA server's faulty hardware from 12/26 was replaced.
+             There were intermittent failures on 12/26 and 1/3. Currently, 
+             there are no known issues. Let us know if you experience any.
+ 2022/10/10: Bugfix: random_seed was not being used for alphafold-multimer.
+             Same structure was returned regardless of defined seed. This
+             has been fixed!
+ 2022/07/13: We have set up a new ColabFold MSA server provided by Korean
+             Bioinformation Center. It provides accelerated MSA generation, 
+             we updated the UniRef30 to 2022_02 and PDB/PDB70 to 220313.

Making Protein folding accessible to all via Google Colab!

Notebooks	monomers	complexes	mmseqs2	jackhmmer	templates
AlphaFold2_mmseqs2	Yes	Yes	Yes	No	Yes
AlphaFold2_batch	Yes	Yes	Yes	No	Yes
RoseTTAFold	Yes	No	Yes	No	No
AlphaFold2 (from Deepmind)	Yes	Yes	No	Yes	No
ESMFold	Yes	Maybe	No	No	No
BETA (in development) notebooks
AlphaFold2_advanced	Yes	Yes	Yes	Yes	No
OmegaFold	Yes	Maybe	No	No	No
OLD retired notebooks
AlphaFold2_complexes	No	Yes	No	No	No
AlphaFold2_jackhmmer	Yes	No	Yes	Yes	No
AlphaFold2_noTemplates_noMD
AlphaFold2_noTemplates_yesMD

FAQ

• Where can I chat with other ColabFold users?
  • See our Discord channel!
• Can I use the models for Molecular Replacement?
  • Yes, but be CAREFUL, the bfactor column is populated with pLDDT confidence values (higher = better). Phenix.phaser expects a "real" bfactor, where (lower = better). See post from Claudia Millán.
• What is the maximum length?
  • Limits depends on free GPU provided by Google-Colab fingers-crossed
  • For GPU: Tesla T4 or Tesla P100 with ~16G the max length is ~1400
  • For GPU: Tesla K80 with ~12G the max length is ~1000
  • To check what GPU you got, open a new code cell and type !nvidia-smi
• Is it okay to use the MMseqs2 MSA server (cf.run_mmseqs2) on a local computer?
  • You can access the server from a local computer if you queries are serial from a single IP. Please do not use multiple computers to query the server.
• Where can I download the databases used by ColabFold?
  • The databases are available at colabfold.mmseqs.com
• I want to render my own images of the predicted structures, how do I color by pLDDT?
  • In pymol for AlphaFold structures: spectrum b, red_yellow_green_cyan_blue, minimum=50, maximum=90
  • If you want to use AlphaFold Colours (credit: Konstantin Korotkov)

    set_color n0, [0.051, 0.341, 0.827]
    set_color n1, [0.416, 0.796, 0.945]
    set_color n2, [0.996, 0.851, 0.212]
    set_color n3, [0.992, 0.490, 0.302]
    color n0, b < 100; color n1, b < 90
    color n2, b < 70;  color n3, b < 50

  • In pymol for RoseTTAFold structures: spectrum b, red_yellow_green_cyan_blue, minimum=0.5, maximum=0.9
• What is the difference between the AlphaFold2_advanced and AlphaFold2_mmseqs2 (_batch) notebook for complex prediction?
  • We currently have two different ways to predict protein complexes: (1) using the AlphaFold2 model with residue index jump and (2) using the AlphaFold2-multimer model. AlphaFold2_advanced supports (1) and AlphaFold2_mmseqs2 (_batch) (2).
• What is the difference between localcolabfold and the pip installable colabfold_batch?
  • LocalColabFold is an installer script designed to make ColabFold functionality available on local users' machines. It supports wide range of operating systems, such as Windows 10 or later (using Windows Subsystem for Linux 2), macOS, and Linux.
• Is there a way to amber-relax structures without having to rerun alphafold/colabfold from scratch?
  • Yes, see this notebook.

Running locally

_Note: Checkout localcolabfold too

It is recommended that you create a conda environment with python version 3.7. If you use a newer python version, you might run into problems when installing tensorflow, as the required version may not be found.

conda create --name my_colabfold python=3.7

Then activate it with activate my_colabfold.

Install ColabFold using the pip commands below. pip will resolve and install all required dependencies and ColabFold should be ready within a few minutes to use. Please check the JAX documentation for how to get JAX to work on your GPU or TPU.

pip install "colabfold[alphafold] @ git+https://github.com/sokrypton/ColabFold"
pip install -q "jax[cuda]>=0.3.8,<0.4" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
# For template-based predictions also install kalign and hhsuite
conda install -c conda-forge -c bioconda kalign2=2.04 hhsuite=3.3.0
# For amber also install openmm and pdbfixer
conda install -c conda-forge openmm=7.5.1 pdbfixer

colabfold_batch <directory_with_fasta_files> <result_dir> 

If no GPU or TPU is present, colabfold_batch can be executed (slowly) using only a CPU with the --cpu parameter.

Generating MSAs for large scale structure/complex predictions

First create a directory for the databases on a disk with sufficient storage (940GB (!)). Depending on where you are, this will take a couple of hours:

./setup_databases.sh /path/to/db_folder

Download and unpack mmseqs (Note: The required features aren't in a release yet, so currently, you need to compile the latest version from source yourself or use a static binary). If mmseqs is not in your PATH, replace mmseqs below with the path to your mmseqs:

# This needs a lot of CPU
colabfold_search input_sequences.fasta /path/to/db_folder msas
# This needs a GPU
colabfold_batch msas predictions

This will create intermediate folder msas that contains all input multiple sequence alignments formated as a3m files and a predictions folder with all predicted pdb,json and png files.

Searches against the ColabFoldDB can be done in two different modes:

(1) Batch searches with many sequences against the ColabFoldDB quires a machine with approx. 128GB RAM. The search should be performed on the same machine that called setup_databases.sh since the database index size is adjusted to the main memory size. To search on computers with less main memory delete the index by removing all .idx files, this will force MMseqs2 to create an index on the fly in memory. MMSeqs2 is optimized for large input sequence sets sizes. For batch searches use the --db-load-mode 0 option.

(2) single query searches require the full index (the .idx files) to be kept in memory. This can be done with e.g. by using vmtouch. Thus, this type of search requires a machine with at least 768GB RAM for the ColabfoldDB. If the index is in memory use to --db-load-mode 3 parameter in colabfold_search to avoid index loading overhead. If they database is already in memory use --db-load-mode 2 option.

Tutorials & Presentations

• ColabFold Tutorial presented at the Boston Protein Design and Modeling Club. [video] [slides].

Projects based on ColabFold or helpers

• Run ColabFold on your local computer by Yoshitaka Moriwaki
• ColabFold/AlphaFold2 for protein structure predictions for Discoba species by Richard John Wheeler
• Cloud-based molecular simulations for everyone by Pablo R. Arantes, Marcelo D. Polêto, Conrado Pedebos and Rodrigo Ligabue-Braun
• getmoonbear is a webserver to predict protein structures by Stephanie Zhang and Neil Deshmukh
• ColabFold/AlphaFold2 IDR complex prediction by Balint Meszaros
• ColabFold/AlphaFold2 (Phenix version) for macromolecular structure determination by Tom Terwilliger
• AlphaPickle: making AlphaFold2/ColabFold outputs interpretable by Matt Arnold

Acknowledgments

• We would like to thank the RoseTTAFold and AlphaFold team for doing an excellent job open sourcing the software.
• Also credit to David Koes for his awesome py3Dmol plugin, without whom these notebooks would be quite boring!
• A colab by Sergey Ovchinnikov (@sokrypton), Milot Mirdita (@milot_mirdita) and Martin Steinegger (@thesteinegger).

How do I reference this work?

• Mirdita M, Schütze K, Moriwaki Y, Heo L, Ovchinnikov S and Steinegger M. ColabFold: Making protein folding accessible to all.
  Nature Methods (2022) doi: 10.1038/s41592-022-01488-1
• If you’re using AlphaFold, please also cite:
  Jumper et al. "Highly accurate protein structure prediction with AlphaFold."
  Nature (2021) doi: 10.1038/s41586-021-03819-2
• If you’re using AlphaFold-multimer, please also cite:
  Evans et al. "Protein complex prediction with AlphaFold-Multimer."
  biorxiv (2021) doi: 10.1101/2021.10.04.463034v1
• If you are using RoseTTAFold, please also cite:
  Minkyung et al. "Accurate prediction of protein structures and interactions using a three-track neural network."
  Science (2021) doi: 10.1126/science.abj8754

---

OLD Updates

  11Mar2022: We use in default AlphaFold-multimer-v2 weights for complex modeling. 
             We also offer the old complex modes "AlphaFold-ptm" or "AlphaFold-multimer-v1"
  04Mar2022: ColabFold now uses a much more powerful server for MSAs and searches through the ColabFoldDB instead of BFD/MGnify. 
             Please let us know if you observe any issues.
  26Jan2022: AlphaFold2_mmseqs2, AlphaFold2_batch and colabfold_batch's multimer complexes predictions are 
             now in default reranked by iptmscore*0.8+ptmscore*0.2 instead of ptmscore
  16Aug2021: WARNING - MMseqs2 API is undergoing upgrade, you may see error messages.
  17Aug2021: If you see any errors, please report them.
  17Aug2021: We are still debugging the MSA generation procedure...
  20Aug2021: WARNING - MMseqs2 API is undergoing upgrade, you may see error messages.
             To avoid Google Colab from crashing, for large MSA we did -diff 1000 to get 
             1K most diverse sequences. This caused some large MSA to degrade in quality,
             as sequences close to query were being merged to single representive.
             We are working on updating the server (today) to fix this, by making sure
             that both diverse and sequences close to query are included in the final MSA.
             We'll post update here when update is complete.
  21Aug2021  The MSA issues should now be resolved! Please report any errors you see.
             In short, to reduce MSA size we filter (qsc > 0.8, id > 0.95) and take 3K
             most diverse sequences at different qid (sequence identity to query) intervals 
             and merge them. More specifically 3K sequences at qid at (0→0.2),(0.2→0.4),
             (0.4→0.6),(0.6→0.8) and (0.8→1). If you submitted your sequence between
             16Aug2021 and 20Aug2021, we recommend submitting again for best results!
  21Aug2021  The use_templates option in AlphaFold2_mmseqs2 is not properly working. We are
             working on fixing this. If you are not using templates, this does not affect the
             the results. Other notebooks that do not use_templates are unaffected.
  21Aug2021  The templates issue is resolved!
  11Nov2021  [AlphaFold2_mmseqs2] now uses Alphafold-multimer for complex (homo/hetero-oligomer) modeling.
             Use [AlphaFold2_advanced] notebook for the old complex prediction logic. 
  11Nov2021  ColabFold can be installed locally using pip!
  14Nov2021  Template based predictions works again in the Alphafold2_mmseqs2 notebook.
  14Nov2021  WARNING "Single-sequence" mode in AlphaFold2_mmseqs2 and AlphaFold2_batch was broken 
             starting 11Nov2021. The MMseqs2 MSA was being used regardless of selection.
  14Nov2021  "Single-sequence" mode is now fixed.
  20Nov2021  WARNING "AMBER" mode in AlphaFold2_mmseqs2 and AlphaFold2_batch was broken 
             starting 11Nov2021. Unrelaxed proteins were returned instead.
  20Nov2021  "AMBER" is fixed thanks to Kevin Pan

---