NeVAE: A Deep Generative Model for Molecular Graphs

By Bidisha Samanta, Abir De, Gourhari Jana, Pratim Kumar Chattaraj, Niloy Ganguly, Manuel Gomez Rodriguez. 2018

241 citations recorded by the source corpus.

Deep generative models have been praised for their ability to learn smooth latent representation of images, text, and audio, which can then be used to generate new, plausible data. However, current generative models are unable to work with molecular graphs due to their unique characteristics—their underlying structure is not Euclidean or grid-like, they remain isomorphic under permutation of the nodes labels, and they come with a different number of nodes and edges. In this paper, we propose NeVAE, a novel variational autoencoder for molecular graphs, whose encoder and decoder are specially designed to account for the above properties by means of several technical innovations. In addition, by using masking, the decoder is able to guarantee a set of valid properties in the generated molecules. Experiments reveal that our model can discover plausible, diverse and novel molecules more effectively than several state of the art methods. Moreover, by utilizing Bayesian optimization over the continuous latent representation of molecules our model finds, we can also find molecules that maximize certain desirable properties more effectively than alternatives.

Subfields: Artificial Intelligence, Materials Chemistry, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Materials Chemistry

In Molecular Biology

Citations in the corpus

Learning Multimodal Graph-to-Graph Translation for Molecular Optimization (2018, methodology): “…task through generative modeling (G´omez-Bombarelli et al., 2016; Segler et al., 2017; Kusner et al., 2017; Dai et al., 2018; Jin et al., 2018; Samanta et al., 2018; Li et al., 2018a) and reinforcement learning (Guimaraes et al., 2017; Olivecrona et al., 2017; Popova et al., 2018; You et…”
GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation (2020, background): “A variety of deep generative models have been proposed for molecular graph generation recently (Segler et al., 2017; Olivecrona et al., 2017; Samanta et al., 2018; Neil et al., 2018).”
MolGenSurvey: A Systematic Survey in Machine Learning Models for Molecule Design (2022, methodology): “…RL+GAN REINVENT [146] None SMILES RL+AR MCMG [147] None SMILES RL+AR GEGL [ [151] 2D-Graph 2D-Graph VAE Hier-VAE [152] 2D-Graph 2D-Graph VAE NeVAE [153] 2D-Graph+3D-Geometry 2D-Graph+3D-Geometry VAE Cogmol [154], [155] Data-agnostic Data-agnostic VAE MSO [115] SMILES SMILES VAE+BO…”
Mol-CycleGAN: a generative model for molecular optimization (2019, methodology): “Variational Autoencoder (VAE) [26] in conjunction with SMILES representation has been used to generate novel molecules from the trained continuous latent space [27, 28].”
A Gentle Introduction to Deep Learning for Graphs (2019, methodology): “These models also provide mechanisms to search for molecules with a desired set of chemical properties [69, 101, 80].”
Self-referencing embedded strings (SELFIES): A 100% robust molecular string representation (2019, background): “zation can lead to high validity of the decoded molecules from a VAE [11]. A range of other techniques for generating semantically valid graphs have been presented in the last two years, for example, [30, 31, 12, 32], to name a few. The objective of these works is the demonstration of generative mod”
MolGAN: An implicit generative model for small molecular graphs (2018, methodology): “(Jin et al., 2018) and the NeVAE (Samanta et al., 2018) model.”
All SMILES Variational Autoencoder (2019, methodology): “…VAE (CVAE) [16] (with results reported in [35]), grammar VAE (GVAE) [35], syntax-directed VAE (SD-VAE) [9], junction tree VAE (JT-VAE) [27], NeVAE [56], semisupervised VAE (SSVAE) [28], graph convolutional policy network (GCPN) [70], molecule deep Q-network (MolDQN) [72], and the DeepChem [69]…”