Multimodal Large Language Models for Inverse Molecular Design with Retrosynthetic Planning

TL;DR

Llamole is a novel multimodal large language model that interleaves text and graph generation, enabling advanced molecular inverse design and retrosynthetic planning, outperforming existing models in controllability and planning accuracy.

Contribution

This paper introduces Llamole, the first multimodal LLM capable of integrated text and graph generation for molecular design and retrosynthesis, combining neural networks with search algorithms.

Findings

Llamole outperforms 14 adapted LLMs across 12 metrics.

Llamole enables flexible, controllable molecular generation.

Effective integration of graph neural networks with LLMs for chemical applications.

Abstract

While large language models (LLMs) have integrated images, adapting them to graphs remains challenging, limiting their applications in materials and drug design. This difficulty stems from the need for coherent autoregressive generation across texts and graphs. To address this, we introduce Llamole, the first multimodal LLM capable of interleaved text and graph generation, enabling molecular inverse design with retrosynthetic planning. Llamole integrates a base LLM with the Graph Diffusion Transformer and Graph Neural Networks for multi-conditional molecular generation and reaction inference within texts, while the LLM, with enhanced molecular understanding, flexibly controls activation among the different graph modules. Additionally, Llamole integrates A* search with LLM-based cost functions for efficient retrosynthetic planning. We create benchmarking datasets and conduct extensive experiments to evaluate Llamole against in-context learning and supervised fine-tuning. Llamole significantly outperforms 14 adapted LLMs across 12 metrics for controllable molecular design and retrosynthetic planning.