Automatic chemical design using a data-driven continuous representation of molecules

Abstract

We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in the set of molecules with fewer that nine heavy atoms.

26 pages, 8 figures

References in corpus (1)

• ChemTS: An Efficient Python Library for de novo Molecular Generation

Cited by in corpus (201)

• A Comprehensive Survey on Graph Neural Networks
• DeepDTA: Deep Drug-Target Binding Affinity Prediction
• Deep Reinforcement Learning for De-Novo Drug Design
• Recent Advances and Applications of Deep Learning Methods in Materials Science
• Molecular Transformer - A Model for Uncertainty-Calibrated Chemical Reaction Prediction
• Machine learning for molecular simulation
• Molecular Contrastive Learning of Representations via Graph Neural Networks
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• GuacaMol: Benchmarking Models for De Novo Molecular Design
• Deep learning for molecular design - a review of the state of the art
• Data-driven materials science: status, challenges and perspectives
• The Role of AI in Drug Discovery: Challenges, Opportunities, and Strategies
• Self-Referencing Embedded Strings (SELFIES): A 100% robust molecular string representation
• Big-Data Science in Porous Materials: Materials Genomics and Machine Learning
• ChatGPT Chemistry Assistant for Text Mining and Prediction of MOF Synthesis
• Accelerating Antimicrobial Discovery with Controllable Deep Generative Models and Molecular Dynamics
• Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics
• Optimization of Molecules via Deep Reinforcement Learning
• A Comprehensive Survey on Deep Graph Representation Learning
• Autonomous discovery in the chemical sciences part I: Progress
• Improving Graph Neural Network Expressivity via Subgraph Isomorphism Counting
• Graph Representation Learning in Biomedicine
• SELFIES and the future of molecular string representations
• Differentiable Learning of Quantum Circuit Born Machine
• Autonomous discovery in the chemical sciences part II: Outlook
• Machine Learning Enabled Computational Screening of Inorganic Solid Electrolytes for Dendrite Suppression with Li Metal Anode
• ChemTS: An Efficient Python Library for de novo Molecular Generation
• Transformer-CNN: Fast and Reliable tool for QSAR
• Mol-CycleGAN - a generative model for molecular optimization
• Polymer Informatics: Current Status and Critical Next Steps
• Towards self-driving laboratories: The central role of density functional theory in the AI age
• Training of Quantum Circuits on a Hybrid Quantum Computer
• Bayesian optimization for tuning and selecting hybrid-density functionals
• Conditional molecular design with deep generative models
• The prospects of quantum computing in computational molecular biology
• Unifying the design space and optimizing linear and nonlinear truss metamaterials by generative modeling
• Perspective on integrating machine learning into computational chemistry and materials science
• Improving Chemical Autoencoder Latent Space and Molecular De novo Generation Diversity with Heteroencoders
• TITAN: T Cell Receptor Specificity Prediction with Bimodal Attention Networks
• A Generative Machine Learning-Based Approach for Inverse Design of Multilayer Metasurfaces
• Keeping it Simple: Language Models can learn Complex Molecular Distributions
• Learning to design drug-like molecules in three-dimensional space using deep generative models
• Coarse-Graining Auto-Encoders for Molecular Dynamics
• Molecular graph generation with Graph Neural Networks
• Exploring Chemical Space using Natural Language Processing Methodologies for Drug Discovery
• Generative chemistry: drug discovery with deep learning generative models
• Gryffin: An algorithm for Bayesian optimization of categorical variables informed by expert knowledge
• Bayesian semi-supervised learning for uncertainty-calibrated prediction of molecular properties and active learning
• Quantum Machine Learning for Chemistry and Physics
• Constrained crystals deep convolutional generative adversarial network for the inverse design of crystal structures
• Scaffold-based molecular design using graph generative model
• Regression Transformer: Concurrent sequence regression and generation for molecular language modeling
• Design of metalloproteins and novel protein folds using variational autoencoders
• Computer-Aided Multi-Objective Optimization in Small Molecule Discovery
• Ab-initio quantum chemistry with neural-network wavefunctions
• Machine Learning on Neutron and X-Ray Scattering
• Deep Learning and Knowledge-Based Methods for Computer Aided Molecular Design -- Toward a Unified Approach: State-of-the-Art and Future Directions
• Machine learning and AI-based approaches for bioactive ligand discovery and GPCR-ligand recognition
• Deep Confidence: A Computationally Efficient Framework for Calculating Reliable Errors for Deep Neural Networks
• Privacy-preserving Artificial Intelligence Techniques in Biomedicine
• Review: Deep Learning in Electron Microscopy
• Discovery of Self-Assembling $\pi$-Conjugated Peptides by Active Learning-Directed Coarse-Grained Molecular Simulation
• Materials science in the era of large language models: a perspective
• Scalable Population Synthesis with Deep Generative Modeling
• A Deep Generative Model for Molecule Optimization via One Fragment Modification
• Generative Models for Automatic Chemical Design
• Inverse methods for design of soft materials
• Error-Controlled Exploration of Chemical Reaction Networks with Gaussian Processes
• Computer-inspired Quantum Experiments
• Graph Neural Networks for Molecules
• Improving Molecular Contrastive Learning via Faulty Negative Mitigation and Decomposed Fragment Contrast
• PubChemQC PM6: A dataset of 221 million molecules with optimized molecular geometries and electronic properties
• Chemical diversity in molecular orbital energy predictions with kernel ridge regression
• Autonomous Reaction Network Exploration in Homogeneous and Heterogeneous Catalysis
• Machine learning-assisted design of material properties
• Generative Chemical Transformer: Neural Machine Learning of Molecular Geometric Structures from Chemical Language via Attention
• ChemGrapher: Optical Graph Recognition of Chemical Compounds by Deep Learning
• Optimizing Molecules using Efficient Queries from Property Evaluations
• Molassembler: Molecular graph construction, modification and conformer generation for inorganic and organic molecules
• Associative Learning Mechanism for Drug-Target Interaction Prediction
• Group SELFIES: A Robust Fragment-Based Molecular String Representation
• Scaffold-constrained molecular generation
• Predicting charge density distribution of materials using a local-environment-based graph convolutional network
• A Deep Generative Model for Graph Layout
• Direct prediction of phonon density of states with Euclidean neural networks
• Fast differentiable DNA and protein sequence optimization for molecular design
• Mastering processing-microstructure complexity through the prediction of thin film structure zone diagrams by generative machine learning models
• Temperature-transferable coarse-graining of ionic liquids with dual graph convolutional neural networks
• Forward and Inverse Design of Kirigami via Supervised Autoencoder
• Efficient Construction Method for Phase Diagrams Using Uncertainty Sampling
• Bidirectional Generation of Structure and Properties Through a Single Molecular Foundation Model
• KekuleScope: prediction of cancer cell line sensitivity and compound potency using convolutional neural networks trained on compound images
• Graph Neural Network Based Coarse-Grained Mapping Prediction
• Universal Activation Function For Machine Learning
• De novo Design of Polymer Electrolytes with High Conductivity using GPT-based and Diffusion-based Generative Models
• Extensive deep neural networks for transferring small scale learning to large scale systems
• Efficiently measuring a quantum device using machine learning
• We Should at Least Be Able to Design Molecules That Dock Well
• Hybrid quantum-classical machine learning for generative chemistry and drug design
• El Agente: An Autonomous Agent for Quantum Chemistry
• Using Molecular Embeddings in QSAR Modeling: Does it Make a Difference?
• Deep Molecular Dreaming: Inverse machine learning for de-novo molecular design and interpretability with surjective representations
• Golem: An algorithm for robust experiment and process optimization
• Conformal Prediction Under Feedback Covariate Shift for Biomolecular Design
• A Combined Machine-Learning / Optimization-Based Approach for Inverse Design of Nonuniform Bianisotropic Metasurfaces
• Scalable Uncertainty Quantification for Deep Operator Networks using Randomized Priors
• Difficulty in chirality recognition for Transformer architectures learning chemical structures from string
• Atomistic structure learning
• Inverse Design of Composite Metal Oxide Optical Materials based on Deep Transfer Learning
• Creativity and Machine Learning: A Survey
• Ensuring thermodynamic consistency with invertible coarse-graining
• Utilizing Reinforcement Learning for de novo Drug Design
• Re-evaluating Retrosynthesis Algorithms with Syntheseus
• Accelerating Material Design with the Generative Toolkit for Scientific Discovery
• Graph Machine Learning for Design of High-Octane Fuels
• Machine learning spectral indicators of topology
• Learning from the Density to Correct Total Energy and Forces in First Principle Simulations
• Simulated annealing for optimization of graphs and sequences
• Edge-based sequential graph generation with recurrent neural networks
• Multi-Objective Latent Space Optimization of Generative Molecular Design Models
• AI-guided inverse design and discovery of recyclable vitrimeric polymers
• Variational Autoencoder based Metamodeling for Multi-Objective Topology Optimization of Electrical Machines
• Adversarial Autoencoders with Constant-Curvature Latent Manifolds
• Navigating the Evolution of Two-dimensional Carbon Nitride Research: Integrating Machine Learning into Conventional Approaches
• Analogical discovery of disordered perovskite oxides by crystal structure information hidden in unsupervised material fingerprints
• Tackling the challenge of a huge materials science search space with quantum-inspired annealing
• EspalomaCharge: Machine learning-enabled ultra-fast partial charge assignment
• Magnetic Field Prediction Using Generative Adversarial Networks
• Generative diffusion model for surface structure discovery
• Learning Interpretable Representations of Entanglement in Quantum Optics Experiments using Deep Generative Models
• Accelerated chemical space search using a quantum-inspired cluster expansion approach
• Recent advances in the Self-Referencing Embedding Strings (SELFIES) library
• An Interpretable Machine Learning Model for Deformation of Multi-Walled Carbon Nanotubes
• Controllable Data Generation by Deep Learning: A Review
• Generative Multi-Form Bayesian Optimization
• Crystal structure prediction of materials with high symmetry using differential evolution
• Multi-fidelity Bayesian Optimization: A Review
• Curiosity in exploring chemical space: Intrinsic rewards for deep molecular reinforcement learning
• Efficient Generator of Mathematical Expressions for Symbolic Regression
• Faster and more diverse de novo molecular optimization with double-loop reinforcement learning using augmented SMILES
• Plug & Play Directed Evolution of Proteins with Gradient-based Discrete MCMC
• Spiers Memorial Lecture: How to do impactful research in artificial intelligence for chemistry and materials science
• Application and Assessment of Deep Learning for the Generation of Potential NMDA Receptor Antagonists
• A generative model for molecule generation based on chemical reaction trees
• LIMO: Latent Inceptionism for Targeted Molecule Generation
• Conditional deep generative models as surrogates for spatial field solution reconstruction with quantified uncertainty in Structural Health Monitoring applications
• Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems
• Predicting drug properties with parameter-free machine learning: Pareto-Optimal Embedded Modeling (POEM)
• ChemSpaceAL: An Efficient Active Learning Methodology Applied to Protein-Specific Molecular Generation
• Language models in molecular discovery
• BatmanNet: Bi-branch Masked Graph Transformer Autoencoder for Molecular Representation
• Decoding Structure-Spectrum Relationships with Physically Organized Latent Spaces
• Charge Transfer Database for Bio-molecule Tight Binding Model Derived from Thousands of Proteins
• Automated patent extraction powers generative modeling in focused chemical spaces
• The JARVIS Infrastructure is All You Need for Materials Design
• Adversarial Learned Molecular Graph Inference and Generation
• Materials Informatics: Emergence To Autonomous Discovery In The Age Of AI
• Machine learning-assisted search for novel coagulants: when machine learning can be efficient even if data availability is low
• Feature-Based Interpolation and Geodesics in the Latent Spaces of Generative Models
• Generative model for learning quantum ensemble via optimal transport loss
• Discovery of structure-property relations for molecules via hypothesis-driven active learning over the chemical space
• Co-orchestration of Multiple Instruments to Uncover Structure-Property Relationships in Combinatorial Libraries
• Does Hessian Data Improve the Performance of Machine Learning Potentials?
• Achieving Robustness to Aleatoric Uncertainty with Heteroscedastic Bayesian Optimisation
• GraphXForm: Graph transformer for computer-aided molecular design
• Preliminary results in using Deep Learning to emulate BLOB, a nuclear interaction model
• Data-driven determination of the spin Hamiltonian parameters and their uncertainties: The case of the zigzag-chain compound KCu${}_4$P${}_3$O${}_{12}$
• Space Group Informed Transformer for Crystalline Materials Generation
• Ranking over Regression for Bayesian Optimization and Molecule Selection
• CircuitVAE: Efficient and Scalable Latent Circuit Optimization
• HelixMO: Sample-Efficient Molecular Optimization in Scene-Sensitive Latent Space
• Superconductor discovery in the emerging paradigm of Materials Informatics
• How Evaluation Choices Distort the Outcome of Generative Drug Discovery
• Evolutionary Monte Carlo of QM properties in chemical space: Electrolyte design
• On the Compressive Power of Boolean Threshold Autoencoders
• Accelerated lignocellulosic molecule adsorption structure determination
• Molecular Dynamics and Machine Learning Unlock Possibilities in Beauty Design -- A Perspective
• Prospecting for Pluripotency in Metamaterial Design
• Accelerated scale bridging with sparsely approximated Gaussian learning
• Improving Molecule Generation and Drug Discovery with a Knowledge-enhanced Generative Model
• On the Size and Width of the Decoder of a Boolean Threshold Autoencoder
• Data Fusion of Deep Learned Molecular Embeddings for Property Prediction
• Molecular design method based on novel molecular representation and variational auto-encoder
• TripOptimizer: Generative 3D Shape Optimization and Drag Prediction using Triplane VAE Networks
• Optimizing Discrete Spaces via Expensive Evaluations: A Learning to Search Framework
• Leveraging Active Subspaces to Capture Epistemic Model Uncertainty in Deep Generative Models for Molecular Design
• Reinforcement Fine-Tuning for Materials Design
• VALID-Mol: a Systematic Framework for Validated LLM-Assisted Molecular Design
• Human-level molecular optimization driven by mol-gene evolution
• Quantum Chemistry Driven Molecular Inverse Design with Data-free Reinforcement Learning
• Learning Conditional Invariance through Cycle Consistency
• PAIL: Performance based Adversarial Imitation Learning Engine for Carbon Neutral Optimization
• Latent Space Arc Therapy Optimization
• MolMiner: Toward Controllable, 3D-Aware, Fragment-Based Molecular Design
• Mixed Continuous and Categorical Flow Matching for 3D De Novo Molecule Generation
• Learning Mechanically Driven Emergent Behavior with Message Passing Neural Networks
• Atoms to Processes: The Role of Artificial Intelligence and Machine Learning in Chemical Engineering
• Developing an AI Course for Synthetic Chemistry Students
• A combined quantum-classical method applied to material design: optimization and discovery of photochromic materials for photopharmacology applications
• Crystal structure prediction using age-fitness multi-objective genetic algorithm and coordination number constraints
• High-Dimensional Bayesian Optimization via Random Projection of Manifold Subspaces