Deep3D-DTA: A Tri-Modal Deep Learning Framework for Binding Affinity Prediction Leveraging 3D Structural...

Area of Science:

• Computational chemistry
• Bioinformatics
• Artificial intelligence in drug discovery

Background:

• Drug-target affinity (DTA) prediction is crucial for drug discovery and repurposing.
• Current computational methods often fail to capture complex molecular interactions due to reliance on limited data types (1D sequences or 2D structures).

Purpose of the Study:

• To introduce Deep3D-DTA, a novel tri-modal deep learning framework for accurate DTA prediction.
• To integrate diverse molecular representations (1D sequence, 2D graph, 3D spatial) for enhanced prediction accuracy.

Main Methods:

• Utilized a pre-trained protein language model for encoding 1D amino acid sequences.
• Generated precise 3D structural representations using interatomic distances and bond angles.
• Employed a hybrid feature extraction module combining graph neural networks and multi-head attention for hierarchical pattern learning.

Main Results:

• Deep3D-DTA significantly outperformed state-of-the-art methods on three benchmark datasets (Davis, KIBA, Metz).
• The tri-modal approach effectively captured complex molecular interactions for improved DTA prediction.

Conclusions:

• Deep3D-DTA offers a robust computational tool for accelerating drug discovery.
• Accurate DTA prediction using Deep3D-DTA can potentially reduce drug development costs.