HCat-GNet: Homogeneous Catalyst Graph Neural Network

Overview

HCat-GNet (Homogeneous Catalyst Graph Neural Network) is a cutting-edge, open-source platform designed to facilitate the virtual evaluation and optimization of homogeneous catalysts. Utilizing Graph Neural Networks (GNNs), HCat-GNet predicts the selectivity of homogeneous catalytic reactions based solely on SMILES representations of participant molecules, significantly speeding up the process of ligand optimization in asymmetric catalysis.

Features

• Predictive Accuracy: Delivers highly accurate predictions of enantioselectivity for metal-ligand catalyzed asymmetric reactions.
• Interpretability: Provides insights into how different ligand modifications affect reaction outcomes, enhancing human understanding and guiding experimental design.
• Flexibility: Agnostic to reaction type, capable of handling a variety of catalytic processes without the need for domain-specific adjustments.

Installation

Prerequisites

• Python 3.8 or 3.9
• Pip (Python package installer)

Setup Instructions

1. Clone the Repository:

   git clone https://github.com/EdAguilarB/hcatgnet.git
   
   cd HCat-GNet
   

2. Install Dependencies:

   pip install -r requirements.txt

Usage

To run all experiments as described in our paper

python run_experiments.py

To run the experiments using the CircuS descriptors

python run_experiments.py --descriptors circus

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Main developer: eduardo.aguilar-bejarano@nottingham.ac.uk

Correspongding author: g.figueredo@nottingham.ac.uk