MolGAN: An implicit generative model for small molecular graphs

Abstract

Deep 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 possible 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 heuristics 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 reinforcement learning objective to encourage the generation of molecules with specific desired chemical properties. In experiments on the QM9 chemical database, we demonstrate that our model is capable of generating close to 100% valid compounds. 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, albeit being susceptible to mode collapse. Code at https://github.com/nicola-decao/MolGAN

arXiv commentCode at https://github.com/nicola-decao/MolGAN

References in corpus (21)

• Adam: A Method for Stochastic Optimization
• Continuous control with deep reinforcement learning
• Semi-Supervised Classification with Graph Convolutional Networks
• Geometric deep learning: going beyond Euclidean data
• Categorical Reparameterization with Gumbel-Softmax
• Spectral Networks and Locally Connected Networks on Graphs
• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Learning to Plan Chemical Syntheses
• Improved Training of Wasserstein GANs
• Improved Techniques for Training GANs
• The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
• Junction Tree Variational Autoencoder for Molecular Graph Generation
• Objective-Reinforced Generative Adversarial Networks (ORGAN) for Sequence Generation Models
• Learning Deep Generative Models of Graphs
• NetGAN: Generating Graphs via Random Walks
• GraphGAN: Graph Representation Learning with Generative Adversarial Nets
• Hyperspherical Variational Auto-Encoders
• Modeling Relational Data with Graph Convolutional Networks
• Graphite: Iterative Generative Modeling of Graphs
• Multi-Objective De Novo Drug Design with Conditional Graph Generative Model
• Adversarial Sets for Regularising Neural Link Predictors

Cited by in corpus (149)

• A Comprehensive Survey on Graph Neural Networks
• Graph Neural Networks: A Review of Methods and Applications
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• Deep learning for molecular design - a review of the state of the art
• A Gentle Introduction to Deep Learning for Graphs
• Autonomous discovery in the chemical sciences part II: Outlook
• Object-Centric Learning with Slot Attention
• Conditional molecular design with deep generative models
• GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation
• Learning Discrete Structures for Graph Neural Networks
• Molecular graph generation with Graph Neural Networks
• Exploring Chemical Space using Natural Language Processing Methodologies for Drug Discovery
• Deep Learning on Graphs: A Survey
• Graph Neural Networks: Taxonomy, Advances and Trends
• Learning on Attribute-Missing Graphs
• Deep Learning and Knowledge-Based Methods for Computer Aided Molecular Design -- Toward a Unified Approach: State-of-the-Art and Future Directions
• Hierarchical Generation of Molecular Graphs using Structural Motifs
• Generative Models for Automatic Chemical Design
• Multi-Objective Molecule Generation using Interpretable Substructures
• Machine Learning on Graphs: A Model and Comprehensive Taxonomy
• D-VAE: A Variational Autoencoder for Directed Acyclic Graphs
• Self-Contrastive Learning with Hard Negative Sampling for Self-supervised Point Cloud Learning
• Deeply learning molecular structure-property relationships using attention- and gate-augmented graph convolutional network
• Scaffold-constrained molecular generation
• Drug-Drug Adverse Effect Prediction with Graph Co-Attention
• Accurate Learning of Graph Representations with Graph Multiset Pooling
• A Deep Generative Model for Graph Layout
• Generative AI and Process Systems Engineering: The Next Frontier
• Deep Set Prediction Networks
• A Survey of Deep Learning for Scientific Discovery
• A Two-Step Graph Convolutional Decoder for Molecule Generation
• Constrained Bayesian Optimization for Automatic Chemical Design
• Reinforcement Learning for Generative AI: State of the Art, Opportunities and Open Research Challenges
• BertGCN: Transductive Text Classification by Combining GCN and BERT
• Deep Molecular Dreaming: Inverse machine learning for de-novo molecular design and interpretability with surjective representations
• All SMILES Variational Autoencoder
• A Flexible Generative Framework for Graph-based Semi-supervised Learning
• Spectral-based Graph Convolutional Network for Directed Graphs
• Learning to Extend Molecular Scaffolds with Structural Motifs
• Sample-Efficient Optimization in the Latent Space of Deep Generative Models via Weighted Retraining
• CogMol: Target-Specific and Selective Drug Design for COVID-19 Using Deep Generative Models
• Fast Graph Generation via Spectral Diffusion
• Graph Residual Flow for Molecular Graph Generation
• Adversarial Autoencoders with Constant-Curvature Latent Manifolds
• Hybrid quantum cycle generative adversarial network for small molecule generation
• GraphDF: A Discrete Flow Model for Molecular Graph Generation
• MARS: Markov Molecular Sampling for Multi-objective Drug Discovery
• Hierarchical Graph-to-Graph Translation for Molecules
• Graph Neural Based End-to-end Data Association Framework for Online Multiple-Object Tracking
• A Review of Graph Neural Networks and Their Applications in Power Systems
• Image-Conditioned Graph Generation for Road Network Extraction
• Controllable Data Generation by Deep Learning: A Review
• Reinforcement Learning for Molecular Design Guided by Quantum Mechanics
• A Model to Search for Synthesizable Molecules
• Differentiable Scaffolding Tree for Molecular Optimization
• Potentials and challenges of polymer informatics: exploiting machine learning for polymer design
• Hit and Lead Discovery with Explorative RL and Fragment-based Molecule Generation
• JANUS: Parallel Tempered Genetic Algorithm Guided by Deep Neural Networks for Inverse Molecular Design
• Neural Attribute Machines for Program Generation
• Labeled Graph Generative Adversarial Networks
• SafeCritic: Collision-Aware Trajectory Prediction
• Edge Representation Learning with Hypergraphs
• C5T5: Controllable Generation of Organic Molecules with Transformers
• Uncertainty quantification of molecular property prediction with Bayesian neural networks
• Constructions in combinatorics via neural networks
• Graph Learning Network: A Structure Learning Algorithm
• Quantum Generative Models for Small Molecule Drug Discovery
• Efficient Generation of Structured Objects with Constrained Adversarial Networks
• Graph Coarsening with Neural Networks
• Study of Deep Generative Models for Inorganic Chemical Compositions
• Generative Graphical Inverse Kinematics
• Graph Deconvolutional Generation
• Deep Generation of Heterogeneous Networks
• Deep Extrapolation for Attribute-Enhanced Generation
• Learning a Continuous Representation of 3D Molecular Structures with Deep Generative Models
• Towards Interpretable Sparse Graph Representation Learning with Laplacian Pooling
• Learning Universal Graph Neural Network Embeddings With Aid Of Transfer Learning
• Adversarial Learned Molecular Graph Inference and Generation
• Encoding Robust Representation for Graph Generation
• Brick-by-Brick: Combinatorial Construction with Deep Reinforcement Learning
• Adaptive Generation of Unrestricted Adversarial Inputs
• Barking up the right tree: an approach to search over molecule synthesis DAGs
• DeepScaffold: a comprehensive tool for scaffold-based de novo drug discovery using deep learning
• Symmetry-Aware Actor-Critic for 3D Molecular Design
• Graph Neural Networks with Parallel Neighborhood Aggregations for Graph Classification
• Artificial Intelligence in Drug Discovery: Applications and Techniques
• Application of quantum-inspired generative models to small molecular datasets
• Spatial Graph Attention and Curiosity-driven Policy for Antiviral Drug Discovery
• A novel method for inference of chemical compounds with prescribed topological substructures based on integer programming
• Drug Discovery Approaches using Quantum Machine Learning
• Deep Q-Learning for Directed Acyclic Graph Generation
• Set-to-Sequence Methods in Machine Learning: a Review
• SCGG: A Deep Structure-Conditioned Graph Generative Model
• GraphTune: A Learning-based Graph Generative Model with Tunable Structural Features
• Building LEGO Using Deep Generative Models of Graphs
• ChemoVerse: Manifold traversal of latent spaces for novel molecule discovery
• A Systematic Assessment of Deep Learning Models for Molecule Generation
• Adversarial Graph Disentanglement
• Benchmarking Inverse Optimization Algorithms for Materials Design
• A comprehensive study on the prediction reliability of graph neural networks for virtual screening
• Node Copying: A Random Graph Model for Effective Graph Sampling
• GraphTER: Unsupervised Learning of Graph Transformation Equivariant Representations via Auto-Encoding Node-wise Transformations
• Deterministic Decoding for Discrete Data in Variational Autoencoders
• Physics-Constrained Predictive Molecular Latent Space Discovery with Graph Scattering Variational Autoencoder
• An In-depth Summary of Recent Artificial Intelligence Applications in Drug Design
• Drug-Drug Interaction Prediction with Wasserstein Adversarial Autoencoder-based Knowledge Graph Embeddings
• Simulating Execution Time of Tensor Programs using Graph Neural Networks
• Discrete Object Generation with Reversible Inductive Construction
• The Atlas for the Aspiring Network Scientist
• Signed Network Modeling Based on Structural Balance Theory
• Artificial Intelligence based Autonomous Molecular Design for Medical Therapeutic: A Perspective
• Auto-decoding Graphs
• Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learning
• Deep Generative Modeling in Network Science with Applications to Public Policy Research
• SHADOWCAST: Controllable Graph Generation
• Generating a Doppelganger Graph: Resembling but Distinct
• Neural Variational Inference For Estimating Uncertainty in Knowledge Graph Embeddings
• Weisfeiler-Lehman Embedding for Molecular Graph Neural Networks
• Research and development of MolAICal for drug design via deep learning and classical programming
• KBGN: Knowledge-Bridge Graph Network for Adaptive Vision-Text Reasoning in Visual Dialogue
• Uncertainty quantification of molecular property prediction using Bayesian neural network models
• TGG: Transferable Graph Generation for Zero-shot and Few-shot Learning
• Top-N: Equivariant set and graph generation without exchangeability
• Neural Consciousness Flow
• Partition and Code: learning how to compress graphs
• SGVAE: Sequential Graph Variational Autoencoder
• Cut-Based Graph Learning Networks to Discover Compositional Structure of Sequential Video Data
• Predicting Nanorobot Shapes via Generative Models
• Deep Learning on Attributed Graphs: A Journey from Graphs to Their Embeddings and Back
• On the Power of Edge Independent Graph Models
• Graph Energy-based Model for Substructure Preserving Molecular Design
• Learning Discrete Energy-based Models via Auxiliary-variable Local Exploration
• Game of GANs: Game-Theoretical Models for Generative Adversarial Networks
• Molecular Design Based on Artificial Neural Networks, Integer Programming and Grid Neighbor Search
• GraphGen-Redux: a Fast and Lightweight Recurrent Model for labeled Graph Generation
• Autoencoding Undirected Molecular Graphs With Neural Networks
• HydroNet: Benchmark Tasks for Preserving Intermolecular Interactions and Structural Motifs in Predictive and Generative Models for Molecular Data
• X-GGM: Graph Generative Modeling for Out-of-Distribution Generalization in Visual Question Answering
• Generating the Graph Gestalt: Kernel-Regularized Graph Representation Learning
• TD-GEN: Graph Generation With Tree Decomposition
• Graph Context Encoder: Graph Feature Inpainting for Graph Generation and Self-supervised Pretraining
• A reproducibility study of "Augmenting Genetic Algorithms with Deep Neural Networks for Exploring the Chemical Space"
• Generate Novel Molecules With Target Properties Using Conditional Generative Models
• Learning to Represent and Predict Sets with Deep Neural Networks
• Node Attribute Generation on Graphs
• Generating Multi-type Temporal Sequences to Mitigate Class-imbalanced Problem
• Cyclic Graph Attentive Match Encoder (CGAME): A Novel Neural Network For OD Estimation
• Towards Conditional Generation of Minimal Action Potential Pathways for Molecular Dynamics
• PiNet: Attention Pooling for Graph Classification