Self-Referencing Embedded Strings (SELFIES): A 100% robust molecular string representation

Abstract

The discovery of novel materials and functional molecules can help to solve some of society's most urgent challenges, ranging from efficient energy harvesting and storage to uncovering novel pharmaceutical drug candidates. Traditionally matter engineering -- generally denoted as inverse design -- was based massively on human intuition and high-throughput virtual screening. The last few years have seen the emergence of significant interest in computer-inspired designs based on evolutionary or deep learning methods. The major challenge here is that the standard strings molecular representation SMILES shows substantial weaknesses in that task because large fractions of strings do not correspond to valid molecules. Here, we solve this problem at a fundamental level and introduce SELFIES (SELF-referencIng Embedded Strings), a string-based representation of molecules which is 100\% robust. Every SELFIES string corresponds to a valid molecule, and SELFIES can represent every molecule. SELFIES can be directly applied in arbitrary machine learning models without the adaptation of the models; each of the generated molecule candidates is valid. In our experiments, the model's internal memory stores two orders of magnitude more diverse molecules than a similar test with SMILES. Furthermore, as all molecules are valid, it allows for explanation and interpretation of the internal working of the generative models.

6+3 pages, 6+1 figures

References in corpus (7)

• Automatic chemical design using a data-driven continuous representation of molecules
• Quantum-Chemical Insights from Deep Tensor Neural Networks
• Deep learning for molecular design - a review of the state of the art
• Discovering physical concepts with neural networks
• Automated Search for new Quantum Experiments
• Experimental GHZ Entanglement beyond Qubits
• Constrained Generation of Semantically Valid Graphs via Regularizing Variational Autoencoders

Cited by in corpus (43)

• Recent Advances and Applications of Deep Learning Methods in Materials Science
• Molecular Contrastive Learning of Representations via Graph Neural Networks
• Machine learning for electronically excited states of molecules
• SELFIES and the future of molecular string representations
• Keeping it Simple: Language Models can learn Complex Molecular Distributions
• Regression Transformer: Concurrent sequence regression and generation for molecular language modeling
• Materials science in the era of large language models: a perspective
• Chemoton 2.0: Autonomous Exploration of Chemical Reaction Networks
• Graph Neural Networks for Molecules
• Materials Representation and Transfer Learning for Multi-Property Prediction
• Generative AI and Process Systems Engineering: The Next Frontier
• Hybrid quantum-classical machine learning for generative chemistry and drug design
• Deep Molecular Dreaming: Inverse machine learning for de-novo molecular design and interpretability with surjective representations
• A Survey on Graph Counterfactual Explanations: Definitions, Methods, Evaluation, and Research Challenges
• Graph Machine Learning for Design of High-Octane Fuels
• AI-guided inverse design and discovery of recyclable vitrimeric polymers
• T cell receptor binding prediction: A machine learning revolution
• Operationally meaningful representations of physical systems in neural networks
• Molecular Quantum Circuit Design: A Graph-Based Approach
• Curiosity in exploring chemical space: Intrinsic rewards for deep molecular reinforcement learning
• Molecular Identification from AFM images using the IUPAC Nomenclature and Attribute Multimodal Recurrent Neural Networks
• Ab initio machine learning of phase space averages
• Tutorial: AI-assisted exploration and active design of polymers with high intrinsic thermal conductivity
• Natural Language-Assisted Multi-modal Medication Recommendation
• Coarse-grained crystal graph neural networks for reticular materials design
• Advancing biomolecular understanding and design following human instructions
• Prediction of the morphological evolution of a splashing drop using an encoder-decoder
• Chemical Language Model Linker: blending text and molecules with modular adapters
• Predicting antimicrobial activity of conjugated oligoelectrolyte molecules via machine learning
• Evolutionary Multi-Objective Design of SARS-CoV-2 Protease Inhibitor Candidates
• A Bayesian Flow Network Framework for Chemistry Tasks
• Improving Molecule Generation and Drug Discovery with a Knowledge-enhanced Generative Model
• ChemScraper: Leveraging PDF Graphics Instructions for Molecular Diagram Parsing
• Large Language Models are In-Context Molecule Learners
• Exploring Data-Driven Chemical SMILES Tokenization Approaches to Identify Key Protein-Ligand Binding Moieties
• Tokenization for Molecular Foundation Models
• An Encoder-Decoder Foundation Chemical Language Model for Generative Polymer Design
• Human-level molecular optimization driven by mol-gene evolution
• Calculated state-of-the art results for solvation and ionization energies of thousands of organic molecules relevant to battery design
• Analysis of Robustness of a Large Game Corpus
• Advancing Molecular Graph-Text Pre-training via Fine-grained Alignment
• Digital Surfactant
• The Quantum Ensemble Variational Optimization Algorithm: Applications to Molecular Inverse Design