SmartCADD: An AI-Integrated Drug Designing Platform

Project Overview

SmartCADD is an open-source platform designed to innovate in the field of drug discovery by integrating deep learning, computer-aided drug design (CADD), and quantum mechanics methodologies. Developed in a user-friendly Python framework, SmartCADD aims to provide researchers and developers with powerful tools for virtual screening and drug design.

Current Version: 0.1.0

Version 0.01 Notes:

• Initial version

Requirements

• Python >=3.10
• Conda

Installation

git clone [email protected]:SMU-CATCO/SmartCADD.git
cd SmartCADD

conda install -n base conda-forge::mamba
mamba env create -f conda-environment.yml 
mamba activate smartcadd

pip install -e .

Usage Examples

To get started with SmartCADD, you can run a simple virtual screening process:

from smartcadd.pipeline import BasicCompoundPipeline
from smartcadd.filters import ADMETFilter
from smartcadd.dataset import IterableDataset

# Create a dataset iterator
dataset = IterableDataset(
    root_dir="./data",
    batch_size=10,
)

# Create a pipeline with an ADMET filter
admet_pipeline = BasicCompoundPipeline(
    data_loader=dataset,
    filters=[
        ADMETFilter(
            alert_collection_path="alert_collection.csv",
            output_dir="./results",
            save_results=True,
        ),
    ],
)

# Run ADMET filtering
filtered_results = admet_pipeline.run_filters()

Roadmap

• Version 0.1.0: Initial release

• Version 0.2.0

  • Additional filters:

    • Autodock Vina docking
    • Default PyTorch model wrappers

  • Improved data readers

    • TFRecord/protobuf support
      • reading from protobuf
      • writing to protobuf
    • distributed data loading

  • Optimization

    • improved multiprocessing for subprocess-based modules (ex: xtb and docking)

  • Additional Modules:

    • New XAI Algorithms:
      • XInsight
      • GNNExplainer
      • PGExplainer

  • Improved command line argument handling

    • support cmd line arguments for filters/modules using CLIs (ex: xtb, docking, etc.)

• Version 0.3.0

  • Generative Models:

    • Graph GANs
    • Generative Flow Networks

  • Distributed pipelines

    • distribute work across multiple copies of a filter running on different machines/CPUs

Contribution Guidelines

We welcome contributions to the SmartCADD project! If you're looking to contribute, please start by reading our contribution guidelines in the CONTRIBUTING.md file. For any contributions, ensure you follow our code standards and submit a pull request for review.

License Information

SmartCADD is released under the MIT License. For more details, see the LICENSE file in our repository.

Support and Community

Join our community to get support, discuss new features, and more! Check out the SUPPORT.md file for ways to get in touch, or join our discussions on the GitHub repository.