Message-passing Neural Networks (MPNN) for predicting molecular properties

Introduction

This repository contains the code for training and evaluating message passing neural networks (MPNNs) for predicting molecular properties.

Data

The included data is a subset from Sereina Riniker.

Usage

The notebook run_and_evaluate.ipynb contains the code for training and evaluating the MPNN.

Requirements

The required python packages are

pip install  dgl -f https://data.dgl.ai/wheels/cu117/repo.html
pip install  dglgo -f https://data.dgl.ai/wheels-test/repo.html
pip install dgllife
pip install torch torchvision torchaudio
pip install rdkit

References

Machine Learning of Partial Charges Derived from High-Quality Quantum-Mechanical Calculations

Deep mind

Deep Graph Library