ColabFold

---

New Updates

+ 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

---

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
BETA (in development) notebooks
AlphaFold2_advanced	Yes	Yes	Yes	Yes	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

• 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
  • 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 a command line interface for our advanced notebooks. pip is a command line version of the alphafold_mmseqs2 and alphafold_batch notebook.

Running locally

Install ColabFold using the pip commands below. pip will resolvei 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 --upgrade "jax[cuda]" -f https://storage.googleapis.com/jax-releases/jax_releases.html  # Note: wheels only available on linux.

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

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). 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 search_results
# This just does a bit of IO
colabfold_split_msas search_results msas
# This needs a GPU
colabfold_batch msas predictions

This will create intermediate folders search_results and msas that you can eventually delete, and a predictions folder with all pdb files.

Tutorials & Presentations

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

Projects based on ColabFold

• Run ColabFold (AlphaFold2_advanced) on your local computer by Yoshitaka Moriwaki
• Colab 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
• AlphaFold2 IDR complex prediction by Balint Meszaros

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.
  bioRxiv (2021) doi: 10.1101/2021.08.15.456425
• 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

  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!

---