Many-Shot In-Context Learning for Molecular Inverse Design

TL;DR

This paper introduces a semi-supervised, multi-modal LLM approach that enhances many-shot in-context learning for molecular inverse design, enabling interactive and improved chemical molecule generation.

Contribution

A novel semi-supervised learning method that leverages high-performing LLM generated molecules and experimental data for better molecular design.

Findings

Significant improvement over existing ICL methods in molecular design.

Effective integration of multi-modal LLM for interactive molecule modification.

Accessible approach suitable for scientific use.

Abstract

Large Language Models (LLMs) have demonstrated great performance in few-shot In-Context Learning (ICL) for a variety of generative and discriminative chemical design tasks. The newly expanded context windows of LLMs can further improve ICL capabilities for molecular inverse design and lead optimization. To take full advantage of these capabilities we developed a new semi-supervised learning method that overcomes the lack of experimental data available for many-shot ICL. Our approach involves iterative inclusion of LLM generated molecules with high predicted performance, along with experimental data. We further integrated our method in a multi-modal LLM which allows for the interactive modification of generated molecular structures using text instructions. As we show, the new method greatly improves upon existing ICL methods for molecular design while being accessible and easy to use for scientists.