A modular reimplementation of AF-Cluster for plug-n-play functionality, incorporation into computational workflows, and HPC/HTC deployment through ColabFold.

• Modular Design: Clean separation of MSA generation, clustering, and structure prediction
• Slurm & Apptainer Compatibility: Integration for high-performance and high-throughput computing clusters
• Modular MSA Clustering: Additional clustering methods can be easily swapped in.

• Python 3.8+
• ColabFold (for structure prediction)
• MMseqs2 (optional, for local MSA generation)

1. Clone the repository:

   git clone git@github.com:gelnesr/AFCluster-2.git
   cd AFCluster-2
   

2. Set up ColabFold locally:

   If you do not already have ColabFold installed, please follow the installation instructions from localcolabfold:

   # For Linux
   wget https://raw.githubusercontent.com/YoshitakaMo/localcolabfold/main/install_colabbatch_linux.sh
   bash install_colabbatch_linux.sh

   After setting up ColabFold, set the relative path in the configs/afcluster.yml file. We recommend also setting up a cache directory.

3. For HPC deployment, run the setup script:

   This will automatically set up the enviornment and set the path for ColabFold to your $SCRATCH/tools folder. Edit this line in the .sh script to the appropriate directory.

   bash scripts/env_setup.sh

4. For Apptainer deployment, run the following set of commands.

   First set up a .sif file to your $SCRATCH/containers folder and a cache directory at $SCRATCH/cache. Edit this line in the .sh script to the appropriate directory.

   bash scripts/build_apptainer.sh

   Then, run this which will automatically set up the enviornment and set the path for ColabFold to your $SCRATCH/tools folder. Edit this line in the .sh script to the appropriate directory.

   bash scripts/env_setup.sh

   Then run module load apptainer or module load singularity to initialize an apptainer. You should set the INPUT.fasta in the command below before running:

   apptainer exec --nv \
        --bind "AFCluster-2:/w","$SCRATCH:$SCRATCH" \
        --env XDG_CACHE_HOME="$CACHE" \
        --env MPLCONFIGDIR="$CACHE" \
        "$IMG" bash -lc '
        cd /w
        source afc/bin/activate
        python afcluster.py --input INPUT.fasta
        '

   ---

Run the pipeline on a FASTA file:

python afcluster.py --input sequences.fasta

• --input: Input FASTA file (required)
• --msa: Pre-computed MSA file (optional)

Modify configs/afcluster.yml to adjust clustering parameters:

keyword: "MAIN"           
gap_cutoff: 0.25
random_seed: 42

cluster_method: "dbscan"

dbscan:
  min_samples: 10
  eps_val: null           
  min_eps: 3
  max_eps: 20.0
  eps_step: 0.5 

path_vars:
  PATH: "/path/to/colabfold/bin:$PATH"
  XDG_CACHE_HOME: "/path/to/cache"
  MPLCONFIGDIR: "/path/to/cache"

1. Input Processing: Load FASTA sequences
2. MSA Generation: Create multiple sequence alignments using colabfold_batch or local MMSeqs
3. Sequence Filtering: Remove sequences with high gap content
4. Clustering: Group similar sequences using specified method
5. Structure Prediction: Run ColabFold on each cluster
6. Output: Generate cluster-specific A3M files and predicted structures with corresponding json/png files

output/
├── sequence_id/
│   ├── sequence_id.a3m          # Generated MSA
│   ├── clusters/
│   │   ├── sequence_id_000.a3m  # Cluster 0
│   │   ├── sequence_id_001.a3m  # Cluster 1
│   │   └── ...
│   └── preds/
│       ├── sequence_id_000/
│       │   ├── s0/              # Structure prediction seed 0
│       │   ├── s1/              # Structure prediction seed 1
│       │   └── ...
│       └── ...

This project is based on the original AF-Cluster implementation.

If you use AFCluster-2 in your research, please cite the following works and acknowledge this implementation:

@article{AFCluster,
  title={Predicting multiple conformations via sequence clustering and AlphaFold2},
  DOI={10.1038/s41586-023-06832-9},
  journal={Nature},
  author={Wayment-Steele, Hannah K. and Ojoawo, Adedolapo and Otten, Renee and Apitz, Julia M. and Pitsawong, Warintra and Hömberger, Marc and Ovchinnikov, Sergey and Colwell, Lucy and Kern, Dorothee},
  year={2023},
}

This software builds upon the following tools and methods:

AF-Cluster - Wayment-Steele, H.K., Ojoawo, A., Otten, R. et al. Predicting multiple conformations via sequence clustering and AlphaFold2. Nature 625, 832–839 (2024). https://doi.org/10.1038/s41586-023-06832-9
[GitHub]

ColabFold - Mirdita, M., Schütze, K., Moriwaki, Y. et al. ColabFold: making protein folding accessible to all. Nat Methods 19, 679–682 (2022). https://doi.org/10.1038/s41592-022-01488-1
[GitHub] | [Local Installation]

MMseqs2 - Steinegger, M. & Söding, J. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nat Biotechnol 35, 1026–1028 (2017). https://www.nature.com/articles/nbt.3988 [GitHub]

AlphaFold - Jumper, J., Evans, R., Pritzel, A. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021). https://doi.org/10.1038/s41586-021-03819-2