MultiscaleDTA: A multiscale-based method with a self-attention mechanism for drug-target binding affinity prediction

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target affinity (DTA) prediction is vital for in silico drug discovery.
• Machine learning methods have advanced DTA prediction.
• Existing convolutional neural networks (CNNs) often overlook multi-scale feature information.

Purpose of the Study:

• To propose a novel deep learning framework, MultiscaleDTA, for accurate drug-target binding affinity prediction.
• To address the limitations of existing CNN-based methods in capturing multi-scale features for DTA.
• To enhance the characterization of drug and target properties through comprehensive feature extraction.

Main Methods:

• Developed an end-to-end deep learning framework named MultiscaleDTA.
• Incorporated multi-scale convolutional neural networks (CNNs) to extract features at different scales.
• Integrated a self-attention mechanism to weigh feature contributions for improved DTA prediction.

Main Results:

• MultiscaleDTA demonstrated competitive performance on both regression and binary classification tasks.
• The framework effectively captures multi-scale and comprehensive features.
• Experimental results show improved accuracy compared to state-of-the-art methods.

Conclusions:

• MultiscaleDTA offers a promising approach for accurate drug-target affinity prediction.
• The integration of multi-scale CNNs and self-attention enhances feature representation.
• This framework can accelerate early-stage drug discovery and development.