license: mit language:
- en metrics:
- code_eval pipeline_tag: graph-ml tags:
- code
- machine_learning
- pytorch
- gat
GAT Molecular logP Predictor
A Graph Attention Network (GAT) model for predicting octanol/water partition coefficients (logP) of small molecules.
Model Details
Model Description
- Architecture: Graph Attention Network (GATv2)
- Input: Molecular graph (atoms as nodes, bonds as edges)
- Output: Predicted logP value (regression)
- Purpose: Predict hydrophobicity of organic compounds
Key Features:
- Processes molecules directly as graphs
- Uses atom and bond features for accurate predictions
- Suitable for drug discovery applications
How to Use
Installation
pip install torch torch-geometric rdkit-pypi