M$^{3}$-20M: A Large-Scale Multi-Modal Molecule Dataset for AI-driven Drug Design and Discovery

TL;DR

M3-20M introduces the largest open-access multi-modal molecule dataset for AI-driven drug design, integrating SMILES, 2D graphs, 3D structures, physicochemical properties, and textual descriptions for over 20 million molecules. The dataset is constructed from PubChem, ZINC, and QM9 and augmented with GPT-3.5-generated descriptions, quality-checked by human experts to ensure scientific accuracy. Seven multi-modal downstream datasets are created to support molecule generation and property prediction, and the authors demonstrate substantial performance gains for both generation and prediction tasks when using M3-20M with FP-ICL and LoRA adaptations across various LLMs. This resource enables more diverse, accurate, and scalable training and fine-tuning for AI-driven drug design and discovery, with public availability facilitating broader adoption and benchmarking.

Abstract

This paper introduces M$^{3}$-20M, a large-scale Multi-Modal Molecule dataset that contains over 20 million molecules, with the data mainly being integrated from existing databases and partially generated by large language models. Designed to support AI-driven drug design and discovery, M$^{3}$-20M is 71 times more in the number of molecules than the largest existing dataset, providing an unprecedented scale that can highly benefit the training or fine-tuning of models, including large language models for drug design and discovery tasks. This dataset integrates one-dimensional SMILES, two-dimensional molecular graphs, three-dimensional molecular structures, physicochemical properties, and textual descriptions collected through web crawling and generated using GPT-3.5, offering a comprehensive view of each molecule. To demonstrate the power of M$^{3}$-20M in drug design and discovery, we conduct extensive experiments on two key tasks: molecule generation and molecular property prediction, using large language models including GLM4, GPT-3.5, GPT-4, and Llama3-8b. Our experimental results show that M$^{3}$-20M can significantly boost model performance in both tasks. Specifically, it enables the models to generate more diverse and valid molecular structures and achieve higher property prediction accuracy than existing single-modal datasets, which validates the value and potential of M$^{3}$-20M in supporting AI-driven drug design and discovery. The dataset is available at https://github.com/bz99bz/M-3.

Figures (6)

• Figure 1: Some examples from our M$^{3}$-20M dataset.
• Figure 2: The construction pipeline of our multi-modal molecule dataset M$^{3}$-20M.
• Figure 3: Molecular text description generation by GPT-3.5.
• Figure 4: Paradigms of model training and finetuning using our dataset.
• Figure 5: 20 most-frequent words in textual descriptions.