Glydentify: A Deep Learning Tool for Glycosyltransferase Classification

Overview

Glydentify is an open-source tool leveraging the power of protein language models for the classification of glycosyltransferases (GTs). This application, grounded in the cutting-edge ESM2 protein language model, offers a novel approach to classifying GTs into fold A families. Glydentify not only achieves high accuracy in prediction but also provides insights into donor binding preferences and key contributing residues, adding an explainable component to its functionality. Glydentify is available on Hugging Face Spaces.

Features

High Accuracy Classification: Classifies GTs into fold A families with 92% accuracy and predicts GT-A donor binding preferences with 88% accuracy.
Explainable AI: Identifies key residues contributing to predictions, offering insights into the underlying mechanisms of enzyme function.
User-Friendly Interface: Built with Gradio, Glydentify offers an intuitive interface requiring no programming experience, making it accessible for a wide range of users.
Web-Based Access: Available directly through any web browser via Hugging Face Spaces.

Components

Training Script: The heart of Glydentify lies in the deep learning model training, contained in family_prediction.ipynb. This Jupyter Notebook includes detailed steps for model training, validation, and evaluation, utilizing PyTorch and the ESM2 model.
Application Folder: The App folder hosts a Python script responsible for loading the trained model and labels, and constructing the Gradio-based web application for easy interaction and function prediction.
Accessible Deployment: Hosted on Hugging Face, the application offers a seamless experience for end-users wanting to explore GT classification without any setup requirements.

How to Use

Model Training: To train your own model or understand the training process, explore family_prediction.ipynb. This notebook requires a fundamental understanding of Python and deep learning.
Running the Application: For end-users, simply visit the Glydentify space on Hugging Face. The web interface allows you to input protein sequences and receive function predictions without any coding requirement.
Local Deployment: Advanced users can set up Glydentify locally by cloning the repository, setting up the environment, and running the application script within the App folder.

Requirements

Python 3.9+
PyTorch
Transformers Library
Gradio

Installation

Clone the repository:

git clone https://github.com/arikat/Glydentify.git
cd Glydentify

Install the required packages from the environment file (requires conda/mamba):

conda env create -f glydentify.yml

Contributing

Contributions to Glydentify are welcome! Whether it's feature enhancements, bug fixes, or documentation improvements, feel free to fork the repository and submit a pull request.

Citation

If you use Glydentify in your research, please cite our accompanying paper:

@article{
  title={Glydentify, a deep learning tool for classifying glycosyltransferase function},
  author={Venkat, A., Zhou, Z., Gill, S. and Kannan, N.},
  journal={manscript in preparation},
  year={2024}}

License

Glydentify is released under the GNU General Public License v3.0 License. See the LICENSE file in the repository for details.