Recent Releases of hcatgnet
hcatgnet - HCat-GNet Software
Overview
HCat-GNet is an advanced graph neural network tool designed for optimizing ligand efficiency in asymmetric catalysis. This release introduces new capabilities for predicting the enantioselectivity of metal-ligand catalyzed reactions using only the SMILES notation of participant molecules, significantly advancing the field of computational chemistry.
Key Features
• **Graph-Based Molecular Representation:** Automatically generates graph representations from SMILES for any asymmetric catalytic reaction.
• **High Interpretability:** Provides insights into chiral ligand substituent effects, enhancing human understanding of structure-selectivity relationships.
• **Performance Enhancements:** Improved prediction accuracy for Rhodium-catalyzed asymmetric 1,4-addition reactions, a critical process in pharmaceutical synthesis.
• **Extensive Benchmarking:** HCat-GNet has been rigorously tested against traditional descriptor-based methods, demonstrating superior or comparable performance.
Updates in This Release
• **Enhanced Learning Algorithms:** Incorporates updated graph convolutional networks that more accurately model complex molecular interactions.
• **Expanded Database Compatibility:** Supports a wider range of datasets, improving the model’s ability to generalize across different chemical spaces.
• **Advanced Error Analysis Tools:** Includes new functionalities for detailed error analysis, helping researchers pinpoint and address prediction inaccuracies.
- Python
Published by EdAguilarB over 1 year ago
hcatgnet - HCat-GNet
This new release includes only the scripts to run the experiments reported in the paper (results and graphs have been removed to minimize memory requirements). We have added functions to analyze the GNN models, calculate the CircuS descriptors, and to create molecular graphs of other datasets, including a BiAryl dataset, Catalytic Asymmetric DeAromatization, and the asymmetric N,S-acetal formation.
- Python
Published by EdAguilarB over 1 year ago