Multi-modality attribute learning-based method for drug-protein interaction prediction based on deep neural network.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug-protein interaction (DPI) prediction is crucial for drug discovery but faces challenges due to limited data and ignored intermolecular information.
• Deep learning methods show promise for DPIs but require improvements in feature representation and data utilization.

Purpose of the Study:

• To develop an innovative deep learning framework, multi-modality attributes (MMA)-DPI, for accurate drug-protein interaction prediction.
• To address limitations of existing DPI methods by incorporating intermolecular information and diverse biological representations.

Main Methods:

• Utilized an augmented transformer module to extract intermolecular sub-structural and chemical semantic information.
• Employed a tri-layer graph convolutional neural network to learn topological features from a heterogeneous biological network.
• Integrated molecular and topological representations using a fully connected neural network and adaptive learning weights for interaction scoring.

Main Results:

• The proposed MMA-DPI framework demonstrated superior performance compared to existing state-of-the-art methods in DPI prediction tasks.
• The integration of multi-modality attributes significantly improved the accuracy and robustness of interaction predictions.
• Experimental evaluations confirmed the effectiveness of the designed framework under various conditions.

Conclusions:

• MMA-DPI offers a powerful and effective approach for drug-protein interaction prediction, advancing drug discovery efforts.
• The framework's ability to leverage multi-modality attributes provides a significant advantage over traditional methods.
• This work highlights the potential of integrating diverse data representations for enhanced predictive modeling in pharmacology.