Introduction Usage Parameters Output Releases

title: ColabFold weight: 40

ColabFold

ModeProteinRNASmall-moleculePTMConstraintspLMMSA serverSplit MSA
ColabFold✅❌❌❌❌❌✅✅

General Usage

ColabFold mode can be run using the command below:

nextflow run nf-core/proteinfold \
    --input samplesheet.csv \
    --outdir <OUTDIR> \
    --mode colabfold \
    --colabfold_db <null (default) | PATH> \
    --colabfold_model_preset "<alphafold2_ptm/alphafold2_multimer_v1/alphafold2_multimer_v2/alphafold2_multimer_v3>" \
    --use_gpu \
    -profile <docker/singularity/podman/shifter/charliecloud/conda/institute>
Warning

--colabfold_model_preset is used to infer how to handle multi-entry fasta files. Choosing alphafold2_ptm will result in a multi-entry fasta being processed as a series of monomer entries rather than as a single oligomeric complex.

By default, --mode colabfold will generate MSA files required for structure prediction using a local execution of the ColabFold search protocol. This protocol uses MMseqs2 to search a uniref30 expandable profile database and construct paired alignments using taxonomic labels. MSAs are enriched with additional unpaired sequences by searching an expandable profile databased of environmental sequences.

Note

Local ColabFold search occurs in a separate module to model inference and the resulting MSA will be cached if downstream modules need to be re-run.

File Structure

The file structure of --colabfold_db must be as follows:

Directory structure 
<colabfold_db>/
├── colabfold_envdb
│   ├── colabfold_envdb_202108_db
│   ├── colabfold_envdb_202108_db_aln
│   ├── colabfold_envdb_202108_db_aln.dbtype
│   └── ...
├── colabfold_uniref30
│   ├── uniref30_2302_db
│   ├── uniref30_2302_db_aln
│   ├── uniref30_2302_db_aln.dbtype
│   └── ...
└── params/
    └── alphafold_params_colab_2022-12-06/
        ├── LICENSE
        ├── params_model_1_multimer_v2.npz
        ├── params_model_1_multimer_v3.npz
        ├── params_model_1.npz
        ├── params_model_2_multimer_v2.npz
        ├── params_model_2_multimer_v3.npz
        ├── params_model_2.npz
        ├── params_model_3_multimer_v2.npz
        ├── params_model_3_multimer_v3.npz
        ├── params_model_3.npz
        ├── params_model_4_multimer_v2.npz
        ├── params_model_4_multimer_v3.npz
        ├── params_model_4.npz
        ├── params_model_5_multimer_v2.npz
        ├── params_model_5_multimer_v3.npz
        └── params_model_5.npz

If individual components are available at different locations in the filesystem, they can be set using the following flags:

--colabfold_envdb_path </PATH/TO/colabfold_envdb/*>
--colabfold_uniref30_path </PATH/TO/colabfold_uniref30/*>
--colabfold_alphafold2_params_path </PATH/TO/params/alphafold_params_colab_2022-12-06/>

Without setting the --colabfold_db flag, all of the required data files will be downloaded during the workflow execution.

Warning

The ColabFold reference sequence databases (uniref30_2302 and colabfold_envdb_202108) require ~1TB of disk space.

As an alternative, ColabFold MSAs can be generated without downloading the large reference sequence databases by calling the public MMSeqs API with the --use_msa_server argument. Users can also point to a private api endpoint using the --msa_server_url argument.

nextflow run nf-core/proteinfold \
    --input samplesheet.csv \
    --outdir <OUTDIR> \
    --mode colabfold \
    --colabfold_db <PATH> \
    --colabfold_model_preset <alphafold2_ptm/alphafold2_multimer_v1/alphafold2_multimer_v2/alphafold2_multimer_v3> \
    --use_msa_server \
    --use_gpu \
    -profile <docker/singularity/podman/shifter/charliecloud/conda/institute>
Warning

If you aim to carry out a large number of predictions, please use the local mmseqs search module or setup and use your own custom MMSeqs2 API Server. You can find instructions here.

Additional Arguments

See the ColabFold documentation for a full description of additional arguments. The arguments supported by the proteinfold workflow are described briefly below:

ParameterDefaultDescription
--colabfold_num_recycles3The AlphaFold2 model used by ColabFold provides initial structure predictions as a recycled model input in an iterative refinement process. This parameter controls the number of times model outputs are recycled. Increasing the number of recycles has been found to improve performance for some challening cases.
--colabfold_use_ambertrueColabFold outputs will sometimes contain phsyical violations such as steric clashes. These clashes can be resolved by post-processing the outputs with a short relaxation using the Amber Force Field. Non-clashing atoms are pinned to starting coordinates such that the relaxation has a minimal impact on final structures.
--colabfold_db_load_mode0Specify the way that MMSeqs2 will load the required databases in memory
--colabfold_use_templatesfalseUse PDB templates to support predictions. The ColabFold notebooks do not use templates by default.
--colabfold_create_indexfalseCreate index for ColabFold databases during setup. On network filesystems it can be more performant to re-compute the index on the fly

You can override any of these parameters via the command line or a params file.
On this page