Molecular Graph Convolutions: Moving Beyond Fingerprints

2016/03/31 by Steven Kearnes, Kevin McCloskey, Marc Berndl +3 · 116 citations
Computer Science · Engineering · Materials Science · Mathematics · #Computational Drug Discovery Methods #Machine Learning in Materials Science #Surface Chemistry and Catalysis #cs.LG #stat.ML

paper · pdf · doi:10.1007/s10822-016-9938-8

published as J Comput Aided Mol Des (2016) · See "Version information" section

arxiv created 2016/08/18 · arxiv updated 2016/08/26

Abstract

Molecular "fingerprints" encoding structural information are the workhorse of cheminformatics and machine learning in drug discovery applications. However, fingerprint representations necessarily emphasize particular aspects of the molecular structure while ignoring others, rather than allowing the model to make data-driven decisions. We describe molecular "graph convolutions", a machine learning architecture for learning from undirected graphs, specifically small molecules. Graph convolutions use a simple encoding of the molecular graph---atoms, bonds, distances, etc.---which allows the model to take greater advantage of information in the graph structure. Although graph convolutions do not outperform all fingerprint-based methods, they (along with other graph-based methods) represent a new paradigm in ligand-based virtual screening with exciting opportunities for future improvement.

Citations

• Iterative partial equalization of orbital electronegativity—a rapid access to atomic charges

Cited by

• Learning Cell-Aware Hierarchical Multi-Modal Representations for Robust Molecular Modeling
• GraphCliff: Short-Long Range Gating for Modeling Critical Activity Changes Caused by Subtle Molecular Differences
• Connecting the concepts of quantum state tomography and molecular representations for machine learning
• Neural Relational Inference for Interacting Systems
• Implicit vs Unfolded Graph Neural Networks
• Tiered Latent Representations and Latent Spaces for Molecular Graphs
• iSKIN: Integrated application of machine learning and Mondrian conformal prediction to detect skin sensitizers in cosmetic raw materials
• Automatic chemical design using a data-driven continuous representation of molecules
• Graph neural networks: A review of methods and applications
• Hierarchical Generation of Molecular Graphs using Structural Motifs
• Deep learning: new computational modelling techniques for genomics
• Symmetry and Generalisation in Neural Approximations of Renormalisation Transformations
• If You Want to Be Robust, Be Wary of Initialization
• Enhancing Graph Classification Robustness with Singular Pooling
• Robustness Verification of Graph Neural Networks Via Lightweight Satisfiability Testing
• Relational Pooling for Graph Representations
• Chemi-net: a graph convolutional network for accurate drug property prediction
• Graph Neural Networks: A Review of Methods and Applications
• Reasoning-Enhanced Large Language Models for Molecular Property Prediction
• Relational Graph Attention Networks
• Hierarchical Graph-to-Graph Translation for Molecules
• When Does Global Attention Help? A Unified Empirical Study on Atomistic Graph Learning
• An End-to-End Framework for Molecular Conformation Generation via\n Bilevel Programming
• Structure-Aware Label Smoothing for Graph Neural Networks
• Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation
• AI‐Guided Inverse Design and Discovery of Recyclable Vitrimeric Polymers
• Graph-Neural-Network-Based Unsupervised Learning of the Temporal Similarity of Structural Features Observed in Molecular Dynamics Simulations
• SB-MTL: Score-based Meta Transfer-Learning for Cross-Domain Few-Shot Learning
• Integrating artificial intelligence in drug discovery and early drug development: a transformative approach
• Functional microRNA-targeting drug discovery by graph-based deep learning
• Quantum Graph Neural Networks
• Message Passing Networks for Molecules with Tetrahedral Chirality
• Deep learning and the Schrödinger equation
• SeaPearl: A Constraint Programming Solver guided by Reinforcement Learning
• A Survey of Deep Learning for Scientific Discovery
• Explainability Techniques for Graph Convolutional Networks
• Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations
• A Graph to Graphs Framework for Retrosynthesis Prediction
• Reliable Graph Neural Networks for Drug Discovery Under Distributional Shift
• Chemical-Reaction-Aware Molecule Representation Learning
• Deep Learning on Attributed Graphs: A Journey from Graphs to Their Embeddings and Back
• Weisfeiler and Leman go Hyperbolic: Learning Distance Preserving Node Representations
• Artificial Intelligence in Drug Discovery: Applications and Techniques
• Supervised Community Detection with Line Graph Neural Networks
• Axiomatic Attribution for Deep Networks
• How Powerful are Graph Neural Networks?
• Symmetry-adapted generation of 3d point sets for the targeted discovery of molecules
• Adversarial Graph Disentanglement
• Recent Advances and Applications of Deep Learning Methods in Materials Science
• Transportation Scenario Planning with Graph Neural Networks
• Transformer Based Molecule Encoding for Property Prediction
• Probabilistic hypergraph grammars for efficient molecular optimization
• Graph Neural Networks Inspired by Classical Iterative Algorithms
• Graphite: Iterative Generative Modeling of Graphs
• Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks
• Discriminative structural graph classification
• Learn molecular representations from large-scale unlabeled molecules for drug discovery
• Graph Neural Tangent Kernel: Fusing Graph Neural Networks with Graph Kernels
• Geometric Multi-color Message Passing Graph Neural Networks for Blood-brain Barrier Permeability Prediction
• Graph Neural Networks for the Prediction of Substrate-Specific Organic Reaction Conditions
• Graph-Convolutional Deep Learning to Identify Optimized Molecular Configurations
• Learning a Continuous Representation of 3D Molecular Structures with Deep Generative Models
• Neural Consciousness Flow
• Attention-Based Learning on Molecular Ensembles
• Prediction of Carbon Nanostructure Mechanical Properties and Role of Defects Using Machine Learning
• IL-Net: Using Expert Knowledge to Guide the Design of Furcated Neural\n Networks
• Prediction of Small Molecule Kinase Inhibitors for Chemotherapy Using Deep Learning
• Graph Convolutional Neural Network-Enabled Frontier Molecular Orbital Prediction: A Case Study with Neurotransmitters and Antidepressants
• Generative quantum combinatorial optimization by means of a novel conditional generative quantum eigensolver
• PADME: A Deep Learning-based Framework for Drug-Target Interaction Prediction
• Particle Track Reconstruction using Geometric Deep Learning
• Generating valid Euclidean distance matrices
• Self-Supervised Graph Transformer on Large-Scale Molecular Data
• LiDAR-based Online 3D Video Object Detection with Graph-based Message Passing and Spatiotemporal Transformer Attention
• Jointly Learning Entity and Relation Representations for Entity Alignment
• Deeply learning molecular structure-property relationships using attention- and gate-augmented graph convolutional network
• All SMILES Variational Autoencoder
• Molecular modeling with machine-learned universal potential functions
• Graph Warp Module: an Auxiliary Module for Boosting the Power of Graph Neural Networks in Molecular Graph Analysis
• Machine learning-accelerated quantum mechanics-based atomistic simulations for industrial applications
• Improving Graph Property Prediction with Generalized Readout Functions
• Deep Learning on Graphs: A Survey
• Scalar Coupling Constant Prediction Using Graph Embedding Local Attention Encoder
• ChemBERTa: Large-Scale Self-Supervised Pretraining for Molecular Property Prediction
• Graph Filtration Learning
• Applications of machine learning in GPCR bioactive ligand discovery
• Molecule3D: A Benchmark for Predicting 3D Geometries from Molecular Graphs
• Gated Graph Recursive Neural Networks for Molecular Property Prediction
• Energy-based View of Retrosynthesis
• ENT-DESC: Entity Description Generation by Exploring Knowledge Graph
• Graph-to-Sequence Learning using Gated Graph Neural Networks
• Generic Vision and Cross-Attention for Reaction Yield Prediction
• Batched Sparse Matrix Multiplication for Accelerating Graph Convolutional Networks
• Molecule Property Prediction and Classification with Graph Hypernetworks
• DeepHealth: Review and challenges of artificial intelligence in health informatics
• Artificial Intelligence based Autonomous Molecular Design for Medical Therapeutic: A Perspective
• Leveraging Partial SMILES Validation Scheme for Enhanced Drug Design in Reinforcement Learning Frameworks
• Making Graph Neural Networks Worth It for Low-Data Molecular Machine Learning
• Spatial statistics for screening molecular structures
• Geometric learning of the conformational dynamics of molecules using dynamic graph neural networks
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
• Demand Forecasting from Spatiotemporal Data with Graph Networks and Temporal-Guided Embedding
• Hierarchical modeling of molecular energies using a deep neural network
• ChemGrapher: Optical Graph Recognition of Chemical Compounds by Deep Learning
• Representation Learning for Natural Language Processing
• Accurate Prediction of Free Solvation Energy of Organic Molecules via Graph Attention Network and Message Passing Neural Network from Pairwise Atomistic Interactions
• Optimizing Diffusion Rate and Label Reliability in a Graph-Based Semi-supervised Classifier
• PushNet: Efficient and Adaptive Neural Message Passing
• Graph neural networks in particle physics
• Symmetrical Graph Neural Network for Quantum Chemistry with Dual Real and Momenta Space
• Molecular distance matrix prediction based on graph convolutional networks
• A Physically-informed Graph-based Order Parameter for the Universal Characterization of Atomic Structures
• LiteGEM: Lite Geometry Enhanced Molecular Representation Learning for Quantum Property Prediction
• Designing compact training sets for data-driven molecular property prediction
• Geometric deep learning for computational mechanics Part I: anisotropic hyperelasticity
• Large language models as uncertainty-calibrated optimizers for experimental discovery
• Using attribution to decode binding mechanism in neural network models for chemistry

Related

• Convolutional Networks on Graphs for Learning Molecular Fingerprints
• MoCL: Data-driven Molecular Fingerprint via Knowledge-aware Contrastive Learning from Molecular Graph
• Graph Representation Learning for Interactive Biomolecule Systems
• Molecular Property Prediction Based on Graph Structure Learning
• Graph Data Modeling: Molecules, Proteins, & Chemical Processes