All SMILES Variational Autoencoder

TL;DR

This paper introduces All SMILES VAE, a novel variational autoencoder that encodes multiple SMILES strings per molecule to create a near-bijective, molecule-based latent space, improving molecular property prediction and optimization.

Contribution

It proposes a new SMILES-based VAE that encodes multiple representations of molecules to enhance the latent space quality and property prediction accuracy.

Findings

Outperforms state-of-the-art in property regression tasks

Achieves near-bijective mapping between molecules and latent space

Improves molecular property optimization results

Abstract

Variational autoencoders (VAEs) defined over SMILES string and graph-based representations of molecules promise to improve the optimization of molecular properties, thereby revolutionizing the pharmaceuticals and materials industries. However, these VAEs are hindered by the non-unique nature of SMILES strings and the computational cost of graph convolutions. To efficiently pass messages along all paths through the molecular graph, we encode multiple SMILES strings of a single molecule using a set of stacked recurrent neural networks, pooling hidden representations of each atom between SMILES representations, and use attentional pooling to build a final fixed-length latent representation. By then decoding to a disjoint set of SMILES strings of the molecule, our All SMILES VAE learns an almost bijective mapping between molecules and latent representations near the high-probability-mass subspace of the prior. Our SMILES-derived but molecule-based latent representations significantly surpass the state-of-the-art in a variety of fully- and semi-supervised property regression and molecular property optimization tasks.