Higher-Order Molecular Grammars for Generative and Foundation Models in Chemistry
What it is
This paper introduces the Higher-order Grammar Representation (HGR), a framework that encodes molecular topology as combinatorial complexes parsed into context-free grammar rule sequences. It also presents RingDiv, a new benchmark of 1.18 million molecules designed to test ring-system diversity, and demonstrates that HGR-based models achieve high validity and performance in both molecular generation and representation learning tasks.
The problem it solves
Standard molecular representations (SMILES, graphs) struggle to explicitly encode higher-order topology like ring systems, and existing higher-order methods are computationally expensive and difficult to decode into valid molecules.
What you could use it for
- Generating valid molecules with complex ring systems
- Training foundation models for molecular property prediction
- Benchmarking molecular generation models on ring diversity
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.