A 3D-Shape Similarity-Based Contrastive Approach to Molecular Representation Learning
Abstract
Molecular shape and geometry dictate key biophysical recognition processes, yet many modern graph neural networks disregard 3D information for molecular property prediction. Here, we propose a new contrastive-learning procedure for graph neural networks, Molecular Contrastive Learning from Shape Similarity (MolCLaSS), that implicitly learns a three-dimensional representation. Rather than directly encoding or targeting three-dimensional poses, MolCLaSS matches a similarity objective based on Gaussian overlays to learn a meaningful representation of molecular shape. We demonstrate how this framework naturally captures key aspects of three-dimensionality that two-dimensional representations cannot and provides an inductive framework for scaffold hopping.
Cite
Text
Atsango et al. "A 3D-Shape Similarity-Based Contrastive Approach to Molecular Representation Learning." NeurIPS 2022 Workshops: AI4Science, 2022.Markdown
[Atsango et al. "A 3D-Shape Similarity-Based Contrastive Approach to Molecular Representation Learning." NeurIPS 2022 Workshops: AI4Science, 2022.](https://mlanthology.org/neuripsw/2022/atsango2022neuripsw-3dshape/)BibTeX
@inproceedings{atsango2022neuripsw-3dshape,
title = {{A 3D-Shape Similarity-Based Contrastive Approach to Molecular Representation Learning}},
author = {Atsango, Austin and Diamant, Nathaniel Lee and Lu, Ziqing and Biancalani, Tommaso and Scalia, Gabriele and Chuang, Kangway V},
booktitle = {NeurIPS 2022 Workshops: AI4Science},
year = {2022},
url = {https://mlanthology.org/neuripsw/2022/atsango2022neuripsw-3dshape/}
}