Generative Chemical Transformer: Neural Machine Learning of Molecular Geometric Structures from Chemical Language via Attention

TL;DR

The paper introduces a neural network model called Generative Chemical Transformer (GCT) that leverages attention mechanisms to generate realistic molecular structures from chemical language, aiding inverse molecular design.

Contribution

It presents a novel attention-based neural network that understands chemical language deeply, enabling the generation of molecules satisfying multiple properties for material discovery.

Findings

GCT generates chemically valid and linguistically consistent molecular strings.

Generated molecules meet multiple target properties simultaneously.

Attention mechanism improves understanding of molecular structures beyond chemical language limitations.

Abstract

Discovering new materials better suited to specific purposes is an important issue in improving the quality of human life. Here, a neural network that creates molecules that meet some desired conditions based on a deep understanding of chemical language is proposed (Generative Chemical Transformer, GCT). The attention mechanism in GCT allows a deeper understanding of molecular structures beyond the limitations of chemical language itself which cause semantic discontinuity by paying attention to characters sparsely. It is investigated that the significance of language models for inverse molecular design problems by quantitatively evaluating the quality of the generated molecules. GCT generates highly realistic chemical strings that satisfy both chemical and linguistic grammar rules. Molecules parsed from generated strings simultaneously satisfy the multiple target properties and vary for a single condition set. These advances will contribute to improving the quality of human life by accelerating the process of desired material discovery.