3DMolNet: A Generative Network for Molecular Structures

TL;DR

This paper introduces 3DMolNet, a variational autoencoder-based generative model that efficiently creates valid 3D molecular structures with high accuracy, overcoming limitations of previous string- or graph-based approaches.

Contribution

It presents a novel VAE model that generates variable-sized, compositionally diverse molecules with invariant representations, achieving state-of-the-art reconstruction accuracy.

Findings

Mean reconstruction error below 0.05 Angstrom

Outperforms existing methods by a factor of four

Generated molecules are validated by quantum chemical methods

Abstract

With the recent advances in machine learning for quantum chemistry, it is now possible to predict the chemical properties of compounds and to generate novel molecules. Existing generative models mostly use a string- or graph-based representation, but the precise three-dimensional coordinates of the atoms are usually not encoded. First attempts in this direction have been proposed, where autoregressive or GAN-based models generate atom coordinates. Those either lack a latent space in the autoregressive setting, such that a smooth exploration of the compound space is not possible, or cannot generalize to varying chemical compositions. We propose a new approach to efficiently generate molecular structures that are not restricted to a fixed size or composition. Our model is based on the variational autoencoder which learns a translation-, rotation-, and permutation-invariant low-dimensional representation of molecules. Our experiments yield a mean reconstruction error below 0.05 Angstrom, outperforming the current state-of-the-art methods by a factor of four, and which is even lower than the spatial quantization error of most chemical descriptors. The compositional and structural validity of newly generated molecules has been confirmed by quantum chemical methods in a set of experiments.