Molecular Notations

String notations for encoding molecular structure, from the original SMILES format through SELFIES and InChI, plus tokenization methods and representation surveys.

Concept References

Paper	Key Idea
SMILES: A Compact Notation for Chemical Structures	Comprehensive overview of SMILES ASCII string notation
SELFIES: A Robust Molecular String Representation	Overview of SELFIES, where every string decodes to a valid molecule
InChI: The International Chemical Identifier	IUPAC’s layered chemical identifier for database interoperability

SMILES Family

Year	Paper	Key Idea
1988	SMILES Notation: The Original Paper by Weininger (1988)	Introduced SMILES encoding rules for molecular strings
2018	DeepSMILES: Adapting SMILES Syntax for Machine Learning	Postfix notation eliminating unbalanced parentheses and ring closures
2019	Randomized SMILES Improve Molecular Generative Models	Non-canonical SMILES augmentation improves RNN generation
2024	Invalid SMILES Benefit Chemical Language Models: A Study	Invalid SMILES generation improves CLMs via quality filtering
2024	t-SMILES: Tree-Based Fragment Molecular Encoding	Fragment-based binary tree traversal reduces nesting depth

SELFIES Family

Year	Paper	Key Idea
2020	SELFIES: The Original Paper on Robust Molecular Strings	100% robust molecular representation for generative ML
2022	SELFIES and the Future of Molecular String Representations	Perspective proposing 16 research directions for SELFIES
2023	Recent Advances in the SELFIES Library: 2023 Update	Streamlined grammar, aromatic support, performance upgrades
2023	Group SELFIES: Fragment-Based Molecular Strings	Fragment group tokens extending SELFIES for distribution learning

InChI Family

Year	Paper	Key Idea
2013	InChI: The Worldwide Chemical Structure Identifier Standard	How InChI became the global standard for chemical identifiers
2018	RInChI: The Reaction International Chemical Identifier	Extends InChI to uniquely identify chemical reactions
2019	Mixfile & MInChI: Machine-Readable Mixture Formats	First standardized machine-readable formats for mixtures
2020	NInChI: Toward a Chemical Identifier for Nanomaterials	Extending InChI to represent multi-component nanomaterials
2020	InChI and Tautomerism: Toward Comprehensive Treatment	86 tautomeric rules validated across 400M+ structures for InChI V2
2025	Making InChI FAIR and Sustainable for Inorganic Chemistry	InChI v1.07 adds inorganic support and FAIR compliance

Tokenization

Year	Paper	Key Idea
2021	SPE: Data-Driven SMILES Substructure Tokenization	BPE-style algorithm learning chemically meaningful SMILES tokens
2023	Atom-in-SMILES: Better Tokens for Chemical Models	Environment-aware atomic tokens reduce token degeneration
2024	SMILES vs SELFIES Tokenization for Chemical LMs	Atom Pair Encoding tokenizer outperforms BPE on both formats
2025	SMI+AIS: Hybridizing SMILES with Environment Tokens	Hybrid SMILES + Atom-In-SMILES tokens improve generation
2026	Smirk: Complete Tokenization for Molecular Models	165-token set achieving full OpenSMILES specification coverage

Surveys and Foundations

Year	Paper	Key Idea
1931	The Number of Isomeric Hydrocarbons of the Methane Series	Recursive formulas for counting alkane structural isomers
2023	Materials Representations for ML Review	Review of solid-state material representations for ML
2025	Review of Molecular Representation Learning Models	Survey of foundation models across five molecular modalities

Computational Chemistry

Caffeine molecular structure with its InChIKey identifier

InChI: The International Chemical Identifier

InChI (International Chemical Identifier) is an open standard from IUPAC that represents molecular structures as hierarchical, layered strings optimized for database interoperability, unique identification, and web search via its hashed InChIKey.

Computational Chemistry

Overview of six categories of materials representations for machine learning

Materials Representations for ML Review

A comprehensive review of how solid-state materials can be numerically represented for machine learning, spanning structural features, graph neural networks, compositional descriptors, transfer learning, and generative models for inverse design.

Computational Chemistry

Bar chart comparing SMILES tokens vs Atom-in-SMILES across molecular generation, retrosynthesis, and reaction prediction

Atom-in-SMILES: Better Tokens for Chemical Models

Introduces Atom-in-SMILES (AIS), a tokenization scheme that encodes local chemical environments into SMILES tokens, improving prediction quality across canonicalization, retrosynthesis, and property prediction tasks.

Computational Chemistry

Bar chart comparing SMILES and DeepSMILES error types, showing DeepSMILES eliminates parenthesis errors

DeepSMILES: Adapting SMILES Syntax for Machine Learning

DeepSMILES replaces paired parentheses and ring closure symbols in SMILES with a postfix notation and single ring-size digits, making it easier for generative models to produce syntactically valid molecular strings.

Computational Chemistry

Bar chart comparing Group SELFIES vs SELFIES on MOSES benchmark metrics

Group SELFIES: Fragment-Based Molecular Strings

Group SELFIES extends SELFIES with group tokens representing functional groups and substructures, maintaining chemical robustness while improving distribution learning and molecular generation quality.

Computational Chemistry

Bar chart showing randomized SMILES generate more of GDB-13 chemical space than canonical SMILES across training set sizes

Randomized SMILES Improve Molecular Generative Models

An extensive benchmark showing that training RNN generative models with randomized (non-canonical) SMILES strings yields more uniform, complete, and closed molecular output domains than canonical SMILES.

Computational Chemistry

Bar chart comparing binding affinity scores across SMILES, AIS, and SMI+AIS hybrid tokenization strategies

SMI+AIS: Hybridizing SMILES with Environment Tokens

Proposes SMI+AIS, a hybrid molecular representation combining standard SMILES tokens with chemical-environment-aware Atom-In-SMILES tokens, demonstrating improved molecular generation for drug design targets.

Computational Chemistry

Bar chart comparing Atom Pair Encoding vs BPE tokenization on MoleculeNet classification tasks

SMILES vs SELFIES Tokenization for Chemical LMs

Introduces Atom Pair Encoding (APE), a chemistry-aware tokenizer for SMILES and SELFIES, and shows it consistently outperforms Byte Pair Encoding in RoBERTa-based molecular property classification on BBBP, HIV, and Tox21 benchmarks.

Computational Chemistry

Visualization of tokenizer vocabulary coverage across chemical space

Smirk: Complete Tokenization for Molecular Models

Introduces Smirk and Smirk-GPE tokenizers that fully cover the OpenSMILES specification, proposes n-gram language models as low-cost proxies for evaluating tokenizer quality, and benchmarks 34 tokenizers across intrinsic and extrinsic metrics.

Computational Chemistry

Bar chart showing SMILES Pair Encoding reduces mean sequence length from 40 to 6 tokens

SPE: Data-Driven SMILES Substructure Tokenization

Introduces SMILES Pair Encoding (SPE), a data-driven tokenization algorithm that learns high-frequency SMILES substrings from ChEMBL to produce shorter, chemically interpretable token sequences for deep learning.

Computational Chemistry

Diagram showing the t-SMILES pipeline from molecular graph fragmentation to binary tree traversal producing a string representation

t-SMILES: Tree-Based Fragment Molecular Encoding

t-SMILES represents molecules by fragmenting them into substructures, building full binary trees, and traversing them breadth-first to produce SMILES-type strings that reduce nesting depth and outperform SMILES, DeepSMILES, and SELFIES on generation benchmarks.

Computational Chemistry

Taxonomy of molecular representation learning foundation models organized by input modality

Review of Molecular Representation Learning Models

A comprehensive survey classifying molecular representation learning foundation models by input modality (sequence, graph, 3D, image, multimodal) and analyzing four pretraining paradigms for drug discovery tasks.

Concept References#

SMILES Family#

SELFIES Family#

InChI Family#

Tokenization#

Surveys and Foundations#

Concept References

SMILES Family

SELFIES Family

InChI Family

Tokenization

Surveys and Foundations