PocketDTA: an advanced multimodal architecture for enhanced prediction of drug−target affinity from 3D structural data of target binding pockets

• DOI: 10.1093/bioinformatics/btae594
• Publication Date: 2024-10-04T16:38:36Z
• Authors: Long Zhao; Hongmei Wang; Shaoping Shi
• Keywords: Interpretability

Abstract Motivation Accurately predicting the drug-target binding affinity (DTA) is crucial to drug discovery and repurposing. Although deep learning has been widely used in this field, it still faces challenges with insufficient generalization performance, inadequate use of three-dimensional (3D) information poor interpretability. Results To alleviate these problems, we developed PocketDTA model. This model enhances performance by pre-trained models ESM-2 GraphMVP. It ingeniously handles first three (top-3) target pockets 3D through customized GVP-GNN Layers GraphMVP-Decoder. Additionally, employs a bilinear attention network enhance Comparative analysis state-of-the-art (SOTA) methods on optimized Davis KIBA datasets reveals that exhibits significant advantages. Further, ablation studies confirm effectiveness components, whereas cold-start experiments illustrate its robust capabilities. In particular, shown advantages identifying key functional groups amino acid residues via molecular docking literature validation, highlighting strong potential for Availability implementation Code data are available at: Https://github.com/zhaolongNCU/PocketDTA.