Improved Drug-target Interaction Prediction with Intermolecular Graph Transformer

TL;DR

This paper introduces Intermolecular Graph Transformer (IGT), a novel deep learning model that effectively captures intermolecular interactions to improve drug-target interaction prediction, outperforming existing methods and demonstrating practical utility in SARS-CoV-2 drug screening.

Contribution

The paper presents a new Transformer-based architecture that models intermolecular information explicitly, enhancing prediction accuracy and generalization in drug-target interaction tasks.

Findings

IGT outperforms state-of-the-art methods by 9.1% and 20.5% in binding activity and pose prediction.

IGT shows strong generalization to unseen receptor proteins.

Successfully identifies active drugs against SARS-CoV-2 with high validation rate.

Abstract

The identification of active binding drugs for target proteins (termed as drug-target interaction prediction) is the key challenge in virtual screening, which plays an essential role in drug discovery. Although recent deep learning-based approaches achieved better performance than molecular docking, existing models often neglect certain aspects of the intermolecular information, hindering the performance of prediction. We recognize this problem and propose a novel approach named Intermolecular Graph Transformer (IGT) that employs a dedicated attention mechanism to model intermolecular information with a three-way Transformer-based architecture. IGT outperforms state-of-the-art approaches by 9.1% and 20.5% over the second best for binding activity and binding pose prediction respectively, and shows superior generalization ability to unseen receptor proteins. Furthermore, IGT exhibits promising drug screening ability against SARS-CoV-2 by identifying 83.1% active drugs that have been validated by wet-lab experiments with near-native predicted binding poses.