Symphony: Symmetry-Equivariant Point-Centered Spherical Harmonics for 3D Molecule Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Daigavane, Ameya, Kim, Song, Geiger, Mario, Smidt, Tess
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912038541852672
author Daigavane, Ameya
Kim, Song
Geiger, Mario
Smidt, Tess
author_facet Daigavane, Ameya
Kim, Song
Geiger, Mario
Smidt, Tess
contents We present Symphony, an $E(3)$-equivariant autoregressive generative model for 3D molecular geometries that iteratively builds a molecule from molecular fragments. Existing autoregressive models such as G-SchNet and G-SphereNet for molecules utilize rotationally invariant features to respect the 3D symmetries of molecules. In contrast, Symphony uses message-passing with higher-degree $E(3)$-equivariant features. This allows a novel representation of probability distributions via spherical harmonic signals to efficiently model the 3D geometry of molecules. We show that Symphony is able to accurately generate small molecules from the QM9 dataset, outperforming existing autoregressive models and approaching the performance of diffusion models.
format Preprint
id arxiv_https___arxiv_org_abs_2311_16199
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Symphony: Symmetry-Equivariant Point-Centered Spherical Harmonics for 3D Molecule Generation
Daigavane, Ameya
Kim, Song
Geiger, Mario
Smidt, Tess
Machine Learning
Biomolecules
We present Symphony, an $E(3)$-equivariant autoregressive generative model for 3D molecular geometries that iteratively builds a molecule from molecular fragments. Existing autoregressive models such as G-SchNet and G-SphereNet for molecules utilize rotationally invariant features to respect the 3D symmetries of molecules. In contrast, Symphony uses message-passing with higher-degree $E(3)$-equivariant features. This allows a novel representation of probability distributions via spherical harmonic signals to efficiently model the 3D geometry of molecules. We show that Symphony is able to accurately generate small molecules from the QM9 dataset, outperforming existing autoregressive models and approaching the performance of diffusion models.
title Symphony: Symmetry-Equivariant Point-Centered Spherical Harmonics for 3D Molecule Generation
topic Machine Learning
Biomolecules
url https://arxiv.org/abs/2311.16199