MolGAN: An implicit generative model for small molecular graphs

By Nicola De Cao, Thomas Kipf. 2018

1,103 citations recorded by the source corpus.

eep generative models for graph-structured data offer a new angle on the problem of chemical synthesis: by optimizing differentiable models that directly generate molecular graphs, it is pos-sible to side-step expensive search procedures in the discrete and vast space of chemical structures. We introduce MolGAN, an implicit, likelihood-free generative model for small molecular graphs that circumvents the need for expensive graph matching procedures or node ordering heuris-tics of previous likelihood-based methods. Our method adapts generative adversarial networks (GANs) to operate directly on graph-structured data. We combine our approach with a reinforce-ment learning objective to encourage the genera-tion of molecules with specific desired chemical properties. In experiments on the QM9 chemi-cal database, we demonstrate that our model is capable of generating close to 100% valid com-pounds. MolGAN compares favorably both to recent proposals that use string-based (SMILES) representations of molecules and to a likelihood-based method that directly generates graphs, al-beit being susceptible to mode collapse.

Subfields: Computational Theory and Mathematics, Materials Chemistry, Molecular Biology, Generative Models, Graph Learning

How researchers see this paper

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

In Computational Theory and Mathematics

In Materials Chemistry

In Molecular Biology

In Generative Models

In Graph Learning

Citations in the corpus

Object-Centric Learning with Slot Attention (2020, background): “Set representations are commonly used in tasks across many data modalities ranging from point cloud prediction [29, 30], classifying multiple objects in an image [31], or generation of molecules with desired properties [32, 33].”
Relational inductive biases, deep learning, and graph networks (2018, background): “Recent work has also focused on building generative models of graphs (Li et al., 2018; De Cao and Kipf, 2018; You et al., 2018; Bojchevski et al., 2018), and unsupervised learning of graph embeddings (Perozzi et al., 2014; Tang et al., 2015; Grover and Leskovec, 2016; Garc´ıa-Dur´an and Niepert,…”
Structure-aware generation of drug-like molecules (2021, methodology): “Early generative methods designed SMILES [7], string-based molecular representations, using language models, RNNs, and GANs [8, 9, 10, 11, 12], discarding molecular graph information.”
MoFlow: An Invertible Flow Model for Generating Molecular Graphs (2020, background): “The MolGAN [6] is a generative adversarial networks (GAN)-based model but shows very limited performance in generating valid and unique molecules.”
Graph Representation Learning (2020, methodology): “In the context of graph generation, the basic GAN approach—first proposed by [De Cao and Kipf, 2018]—is similar to the graph-level VAE discussed in the previous section.”
Integrating Scientific Knowledge with Machine Learning for Engineering and Environmental Systems (2020, background): “There is also recent work on encoding domain knowledge in geometric deep learning [41, 50] that is finding increasing use in computational chemistry [71, 81, 95], physics [25, 54], hydrology [133], geostatistics [14, 294], and neuroscience [155].”
Machine Learning on Graphs: A Model and Comprehensive Taxonomy (2020, neutral): “generating molecules with some desired properties [30, 72, 74, 104, 121].”
Sample-Efficient Optimization in the Latent Space of Deep Generative Models via Weighted Retraining (2020, background): “To realize LSO, g can be chosen to be a state-of-the-art deep generative model (DGM), such as a variational autoencoder (VAE) [35, 55] or a generative adversarial network (GAN) [21], which have been shown to be capable of learning vector representations of many types of high-dimensional, structured ”
Conditional Structure Generation through Graph Variational Generative Adversarial Nets (2019, background): “Moreover, earlier works only model particular families of structures [22, 33], while more recent works mostly consider single graphs with fixed sizes [18, 4, 7, 25, 52, 12].”
Deep Learning with Graph Convolutional Networks: An Overview and Latest Applications in Computational Intelligence (2023, background): “[126] proposed MolGAN, which designs molecular structures containing specifc chemical properties through generative adversarial networks combined with graph convolutional neural networks International Journal of Intelligent Systems 23”
Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations (2022, methodology): “Among them, autoregressive models (You et al., 2018b; Liao et al., 2019) construct a graph via sequential decisions, while one-shot generative models (De Cao & Kipf, 2018; Liu et al., 2019) generate components of a graph at once.”
Integrating Physics-Based Modeling with Machine Learning: A Survey (2020, background): “There is also recent work on encoding domain knowledge for geometric deep learning [34, 41] that is finding increasing use in computational chemistry [61, 70, 83], physics [18, 45], geostatistics [8, 272], and neuroscience [136].”