MOL-Mamba: Enhancing Molecular Representation with Structural & Electronic Insights

TL;DR

MOL-Mamba introduces a novel framework that combines structural and electronic insights to improve molecular representations, leading to better performance in property prediction and drug design tasks.

Contribution

The paper presents MOL-Mamba, a new method integrating structural and electronic information through hierarchical reasoning and semantic fusion, addressing limitations of existing GNN and GT models.

Findings

Outperforms state-of-the-art methods on eleven datasets

Enhances molecular property prediction accuracy

Effectively combines structural and electronic data

Abstract

Molecular representation learning plays a crucial role in various downstream tasks, such as molecular property prediction and drug design. To accurately represent molecules, Graph Neural Networks (GNNs) and Graph Transformers (GTs) have shown potential in the realm of self-supervised pretraining. However, existing approaches often overlook the relationship between molecular structure and electronic information, as well as the internal semantic reasoning within molecules. This omission of fundamental chemical knowledge in graph semantics leads to incomplete molecular representations, missing the integration of structural and electronic data. To address these issues, we introduce MOL-Mamba, a framework that enhances molecular representation by combining structural and electronic insights. MOL-Mamba consists of an Atom & Fragment Mamba-Graph (MG) for hierarchical structural reasoning and a Mamba-Transformer (MT) fuser for integrating molecular structure and electronic correlation learning. Additionally, we propose a Structural Distribution Collaborative Training and E-semantic Fusion Training framework to further enhance molecular representation learning. Extensive experiments demonstrate that MOL-Mamba outperforms state-of-the-art baselines across eleven chemical-biological molecular datasets.