SMILES
A cheminformatics standard for representing molecular structures as ASCII strings, encoding molecules into compact text that is convenient for processing by machine learning models.
Timeline (last 90 days)
SMILES (Simplified Molecular-Input Line-Entry System) is a text format that represents molecular structures as linear strings, generating sequences of atoms and bonds via depth-first traversal of the molecular graph.
Under matched corpus size and compute, randomized SAFE's true yield is 2.8x that of randomized SMILES.
Randomizing only the atom traversal order of SMILES achieves just one-tenth of the genuinely novel molecule yield of SAFE with both axes (atom order and fragment order) randomized simultaneously.
Among molecular representation methods, SMILES strings are compact and can be directly fed to language models, but they are not unique and are insensitive to subtle structural changes; molecular fingerprints (such as ECFP/Morgan fingerprints) are suited to similarity search and traditional machine learning but lose topological details; graph neural networks treat atoms as nodes and chemical bonds as edges and are the current mainstream approach for molecular property prediction and generation.
SMILES has the defect of non-canonicality (the same molecule can have multiple valid representations) and cannot directly encode three-dimensional spatial information.
SMILES is a linear molecular notation proposed by David Weininger in 1988 that encodes the two-dimensional topological structure of a molecule as an ASCII string.
SMILES (Simplified Molecular Input Line Entry System) is a linear molecular notation proposed by David Weininger in 1988.
The same molecule can have multiple valid SMILES representations (i.e., non-canonicality), and the string cannot directly encode three-dimensional spatial information.
All Facts (8)
SMILES (Simplified Molecular-Input Line-Entry System) is a text format that represents molecular structures as linear strings, generating sequences of atoms and bonds via depth-first traversal of the molecular graph.
50%UnverifiedUnder matched corpus size and compute, randomized SAFE's true yield is 2.8x that of randomized SMILES.
50%UnverifiedRandomizing only the atom traversal order of SMILES achieves just one-tenth of the genuinely novel molecule yield of SAFE with both axes (atom order and fragment order) randomized simultaneously.
50%UnverifiedAmong molecular representation methods, SMILES strings are compact and can be directly fed to language models, but they are not unique and are insensitive to subtle structural changes; molecular fingerprints (such as ECFP/Morgan fingerprints) are suited to similarity search and traditional machine learning but lose topological details; graph neural networks treat atoms as nodes and chemical bonds as edges and are the current mainstream approach for molecular property prediction and generation.
50%UnverifiedSMILES has the defect of non-canonicality (the same molecule can have multiple valid representations) and cannot directly encode three-dimensional spatial information.
50%UnverifiedSMILES is a linear molecular notation proposed by David Weininger in 1988 that encodes the two-dimensional topological structure of a molecule as an ASCII string.
50%UnverifiedSMILES (Simplified Molecular Input Line Entry System) is a linear molecular notation proposed by David Weininger in 1988.
50%UnverifiedThe same molecule can have multiple valid SMILES representations (i.e., non-canonicality), and the string cannot directly encode three-dimensional spatial information.
50%