Protein Domain Segmentation Library

This library provides a convenient interface for protein domain segmentation using the powerful tools Merizo and Chainsaw. The library abstracts the complexity of using these tools, allowing users to predict protein domain boundaries directly from PDB files or MDAnalysis universes.

Features

• Unified Interface: Use a consistent API to access both Merizo and Chainsaw functionalities.
• Flexible Input Support: Predict from PDB files or MDAnalysis universes.
• Lightweight Usage: Avoid integrating full tool codebases directly into your project.

Requirements

Python 3.10+

Installation

Install the library and its dependencies using pip:

pip install protein-domain-segmentation

Usage

Below is an example demonstrating how to use the MerizoCluster and ChainsawCluster classes:

from protein_domain_segmentation import MerizoCluster, ChainsawCluster
import MDAnalysis as mda

# Initialize clusters
merizo_cluster = MerizoCluster()
chainsaw_cluster = ChainsawCluster()

# Example PDB file path
pdb_path = "example.pdb"

# Predict chopping using Merizo
merizo_result = merizo_cluster.predict_from_pdb(pdb_path)
print("Merizo Results:", merizo_result)

# Predict chopping using Chainsaw
chainsaw_result = chainsaw_cluster.predict_from_pdb(pdb_path)
print("Chainsaw Results:", chainsaw_result)

# Predict chopping from an MDAnalysis Universe
u = mda.Universe(pdb_path)
merizo_from_universe = merizo_cluster.predict_from_universe(u)
print("Merizo Results from Universe:", merizo_from_universe)

Warning for non-Linux Users (Merizo only)

If you are not on a Linux system, you might need to compile Stride, which is used by Merizo for secondary structure assignment.

You can find Stride installation procedure on their official website.

You can then specify STRIDE_PATH to point towards the binary:

export STRIDE_PATH=/path/to/stride_executable

API Overview

Classes

• Cluster: Abstract base class for clustering tools.

  • predict_from_pdb(pdb_path: str, model_params: Dict = None): Predicts the chopping of a protein from a PDB file.
  • predict_from_universe(universe: mda.Universe, model_params: Dict = None): Predicts the chopping of a protein from an MDAnalysis Universe.

• MerizoCluster: Implementation of Cluster for Merizo.

• ChainsawCluster: Implementation of Cluster for Chainsaw.

Methods

• MerizoCluster.predict_from_pdb(pdb_path: str, model_params: Dict = None): Uses Merizo to predict chopping from a PDB file.

• ChainsawCluster.predict_from_pdb(pdb_path: str, model_params: Dict = None): Uses Chainsaw to predict chopping from a PDB file.

Credits

This package wraps the functionality of the following tools:

• Merizo: Developed by the UCL Bioinformatics Group. Visit the Merizo GitHub repository.
• Chainsaw: Developed by Jude Wells and collaborators. Visit the Chainsaw GitHub repository.

Citation

If you use this package in your research, please cite the respective papers for Merizo and Chainsaw:

• Merizo: Lau et al., 2023. Merizo: a rapid and accurate protein domain segmentation method using invariant point attention. Nature Communications.
• Chainsaw: Wells et al., Chainsaw: protein domain segmentation with fully convolutional neural networks. Bioinformatics.

License

This package is distributed under the MIT License. See the LICENSE file for details.

Acknowledgments

We extend our gratitude to the developers of Merizo and Chainsaw for their invaluable contributions to the field of protein domain segmentation.