Command: openfold

OpenFold is an artificial-intelligence method for predicting biomolecular structures consisting of proteins, RNA, DNA, and other molecules such as ligands, cofactors, and drugs [OpenFold3 documentation]. Based on AlphaFold 3, OpenFold 3 is fully open-source and freely available for both academic and commercial use under the Apache 2.0 license. Any work that cites OpenFold 3 should also cite AlphaFold 3:

Accurate structure prediction of biomolecular interactions with AlphaFold 3. Abramson J, Adler J, et al. Nature. 2024 Jun;630(8016):493-500.

The openfold command is also implemented as the OpenFold tool.

The predicted structures vary in confidence levels (see coloring) and should be interpreted with caution. Residue-residue predicted aligned error (PAE) values can be plotted with alphafold pae. See OpenFold structure prediction in ChimeraX. See also: alphafold, boltz, esmfold, modeller, computational screening for protein-protein interactions

Installing OpenFold
Running an OpenFold Prediction
Reopening Batch Ligand Predictions
Running OpenFold on a Server
Limitations

[back to top: openfold]

Installing OpenFold

Usage: openfold install [ directory ] [ downloadModelWeights  true | false ] branch  branch-name

The openfold install command creates a Python virtual environment to install the ChimeraX OpenFold fork Github. If no directory is specified, then ~/openfold3 in the user's home directory is used. The directory will be created, or if it already exists must be empty. The OpenFold network parameters are downloaded to ~/.openfold3.

The install uses a fork of the OpenFold repository https://github.com/RBVI/openfold-3. It uses git branch chimerax_openfold of this fork unless the branch option is specified, in which case it installs the specified branch.

The ChimeraX Python executable is used to create the virtual environment. If the ChimeraX installation is moved or deleted, OpenFold will need to be reinstalled. It will also stop working if the openfold directory itself is moved since the executable refers to the install location to find python. (Otherwise, the installation need only be done once per computer.)

The following commands are used to make the virtual environment and install OpenFold. On Windows, if Nvidia graphics is detected, a version of torch with CUDA 12.6 support is installed before openfold.

      python -m venv directory
      directory/bin/python -m pip install torch --index-url https://download.pytorch.org/whl/cu126  # On Windows with Nvidia GPU only.
      directory/bin/python -m pip install openfold
      directory/bin/python chimerax/site-packages/openfold/download_weights.py
    

[back to top: openfold]

Running an OpenFold Prediction

Usage: openfold predict [ sequences ] [ protein  sequences ] [ dna  sequences ] [ rna  sequences ] [ ligands  residue-spec  [ excludeLigands  CCD-names ]] [ ligandCcd  CCD-names ] [ ligandSmiles  SMILES-string ] [ forEachSmilesLigand  name,SMILES-string,name,SMILES-string... ] [ name  prediction-name ] [ resultsDirectory  directory ] [ device  default | cpu | gpu ] [ precision  32-true | 16-true | bf16-true | bf16-mixed ] [ samples  N ] [ seed  K ] [ useMsaCache  true  false ] [ msaOnly  true  false ] [ open  true  false ] [ installLocation  directory ] [ wait  true  false ] [ useServer  true  false [ serverHost  server-hostname ][ serverPort  port-number ]]

Biopolymer chains.
The sequences of biopolymer chains to predict can be given as a comma-separated list of any of the following:

1. a chain-spec for one or more chains in atomic structure(s) open in ChimeraX
2. the sequence-spec of a sequence in the Sequence Viewer, in the form:  alignment-ID:sequence-ID  (details...)
3. a UniProt name or accession number for a protein chain
4. a plain-text string of 1-letter residue codes pasted directly into the command line

If given with the protein, dna, or rna keywords, the sequences argument can be of the same form as described above, but chains other than protein, DNA, or RNA (respectively) will be excluded. The dna option will interpret single-letter codes as DNA, the rna option will interpret single-letter codes as RNA, and neither will accept UniProt identifiers since they are only for protein chains. The protein, dna, and rna options can be used more than once in the same command.

Ligand, cofactor, and ion components.
Residues present in currently open structures can be specified with ligands residue-spec, optionally with excludeLigands to omit specific types from that set. For example, if ligands #1 was given but not all of the small molecules in #1 are wanted, excludeLigands would be used to list which residue types to omit. Residues to exclude are specified by a comma-separated list of their 3- or 5-letter residue names in the PDB Chemical Component Dictionary (CCD). By default, CCD name HOH (water) is excluded. Ligands to include can also be specified by a comma-separated list of CCD names with the ligandCcd option, or by a comma-separated list of SMILES strings with the ligandSmiles option. The ligandCcd and ligandSmiles options can be used more than once in the same command.

Batch predictions of different ligands with the same macromolecular chains.
The forEachSmilesLigand option is for batch prediction of the structures of different small-molecule ligands in complex with the same set of biopolymeric components (protein and/or nucleic acid chains). The ligands are specified as a comma-separated list of names and SMILES strings.

Calculation options:

• name  prediction-name
  – a name for the prediction to be used in naming the output folder and files
• resultsDirectory  directory
  – the pathname (name and location) of a folder or directory in which to store prediction results. It may include “[name]” to indicate substitution by the specified prediction-name. If the folder already exists, a numeric suffix _1, _2, _3... will be appended.
• device  default | cpu | gpu
  – whether to run the computation on the GPU or CPU. The default setting chooses based on the availability of an Nvidia or Mac M series GPU and Torch support for the GPU.
• precision  32-true | 16-true | bf16-true | bf16-mixed
  – precision of floating-point operations by PyTorch_Lightning, the machine-learning toolkit used by OpenFold:
  • 32-true – 32-bit IEEE floating point
  • 16-true – 16-bit IEEE floating point
  • bf16-true – always use bfloat16 (bf16), a different format of 16-bit floating point that maintains the dynamic range of 32-bit (1e37 max value) and sacrifices digits only, about 3 digits of precision
  • bf16-mixed – use bfloat16 for some parts of the calculation, 32 for others that typically require a higher precision
  The default is bf16-mixed for predictions using Nvidia GPUs with CUDA, otherwise 32-true.
• samples  N
  – number of predictions (default 1). This is what OpenFold calls "diffusion samples." Creating additional structures takes much less time than creating the first structure.
• seed  K
  – random number seed (an integer) to initialize the calculation. Runs with different seeds will give different results.
• useMsaCache  true | false
  – whether to cache (and potentially reuse) the deep sequence alignments generated by the Colabfold server for protein chains (default true). The alignment cache location is ~/Downloads/ChimeraX/OpenFoldMSA/. Reusing an alignment saves time when multiple predictions will be performed for the same protein or set of proteins but different small-molecule ligands. Because the alignments for different proteins in an assembly are paired to match ones from the same organisms, the cached alignments can only be reused for assemblies with the exact same set of proteins. Alignments computed for individual proteins from multiple different runs cannot be used for an assembly of those proteins.
• msaOnly true|false
  – whether to just calculate and cache the multiple sequence alignment using the Colabfold server. This allows computing the MSA separately from the structure. It can be useful to precompute and cache MSAs when running structure predictions on a cluster that does not have internet access.
• open  true | false
  – whether to open the predicted structures when the prediction finishes (default true). Initial coloring is by residue confidence values (details...). The structures will be aligned to an already open model with matchmaker if that open model was the first used in specifying the assembly.
• installLocation  directory
  – where OpenFold is installed. If specified, this sets the default location for future ChimeraX sessions.
• wait  true | false
  – whether the calculation should freeze ChimeraX until it finishes or allow ChimeraX use during the computation (default, wait false).
• useServer  true | false [ serverHost  server-hostname ][ serverPort  port-number ] – whether to run OpenFold on a server instead of on the local system (default false); if true, the other options specify the server hostname (or IP address) and port number. The OpenFold server must have been set up already (details...).

[back to top: openfold]

Reopening Batch Ligand Predictions

Usage: openfold ligandtable  runDirectory  [ alignTo  chain-spec ]

The openfold ligandtable command opens a previously calculated set of batch ligand predictions and shows them in a table. The results are found under runDirectory (where OpenFold was run). This directory can be specified as a pathname or the word browse to specify it interactively in a file browser window. The alignTo option indicates that any results opened with the table's Open button should be superimposed on the specified structure chain.

[back to top: openfold]

Running OpenFold on a Server

Usage: openfold server start [ host  hostname ] [ port  port-number ] [ openfoldExe  executable-path ] [ jobsDirectory  directory ] [ device  gpu | cpu ] [ gpus  gpu-ids ] [ extraOptions  options ]
Usage: openfold server list
Usage: openfold server fetch [ local-directory ] [ open  true | false ]

Running OpenFold on a server (more powerful than the local machine) allows for faster predictions and enables predicting larger structures. There are options in the OpenFold tool and the openfold predict command to use the server. A user can quit from ChimeraX while the server job is still running, then retrieve the results later (after resuming use of ChimeraX) from the OpenFold History tool or with the openfold server fetch command. See also: OpenFold Prediction Server

The server can be Linux, Windows, or Mac. The service must be set up by installing ChimeraX (from February 13, 2026 or later) on the server computer and running the following commands.

      chimerax --nogui
      > openfold install
      > openfold server start
      > exit

The example above starts ChimeraX in text-only mode, but the openfold commands could just as well be entered into the ChimeraX GUI. Exiting from ChimeraX leaves the server running. The server listens for prediction requests from ChimeraX from other computers, runs the predictions, and returns the results. By default, the server saves jobs files in the directory: ~/openfold_server_jobs

Options for openfold server start:

• host  hostname – server hostname or IP address, for example, minsky.cgl.ucsf.edu or 169.230.21.20 (default is to use the python gethostbyname() system call to get the host name or IP address)
• port port-number – port number on which to listen for connections (default 30172)
• openfoldExe  executable-path – path to the openfold executable (default is to use the install location, which is set initially by openfold install, but can be changed later in the OpenFold options or with the openfold predict command installLocation option
• jobsDirectory  directory – directory in which to create server job subdirectories (default ~/openfold_server_jobs). A text file named openfold_server_log is generated within this directory when the server is started and can be used to debug server errors. The directory will be created if it does not exist.
• device  gpu | cpu – whether the server should run OpenFold using the GPU (default) or the CPU
• gpus  gpu-ids – use multiple GPUs, specified as a comma-separated list of integer GPU IDs. This allows the server to run multiple prediction jobs at the same time, one per GPU. The default is to use only one GPU, even if the computer has multiple GPUs.
• extraOptions  options – use the specified openfold command-line options for every prediction. For instance, the low-memory OpenFold options “--use_cpu_memory --inplace_operations” can allow predicting larger structures.

The openfold server list command lists in the Log any active jobs, defined as those which were still running when the user quit from ChimeraX. The job identifiers, server hostname, and port numbers are listed, along with job status (finished, running, or waiting to run).

The openfold server fetch command retrieves the results of finished jobs. Options for openfold server fetch:

• local-directory – openfold job directory on the local computer. This directory can be specified as a pathname or the word browse to specify it interactively in a file browser window. If the directory is not specified, results for all active jobs that have finished will be fetched. Active jobs are those which were still running when the user quit from ChimeraX. If a job is still running or waiting to run, this status will be reported in the Log.
• open  true | false – whether to open the predicted structures for finished jobs (default true; if false, they can be opened later using the OpenFold History)

There is currently no command to stop the OpenFold server. On Linux or Mac, the process ID can be determined and the process killed with shell commands:

      $ ps -axww | grep openfold
        34435 ChimeraX.app/bin/python3.11 ChimeraX.app/lib/python3.11/site-packages/chimerax/openfold/server.py ...
      $ kill 34435

On Windows, the Task Manager can be used to find the openfold server process and force quit from it. Similarly, on Mac, the Activity Monitor can be used to find the openfold server process and force quit from it.