Modality-DTA: Multimodality Fusion Strategy for Drug-Target Affinity Prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Drug-target affinity (DTA) prediction is crucial for efficient drug discovery.
• Current deep learning methods often rely on single data types (e.g., simplified molecular input line entry specification (SMILES) or amino acid sequences) for DTA prediction.
• Multimodality data offer complementary information for more robust DTA prediction.

Purpose of the Study:

• To introduce Modality-DTA, a novel deep learning framework that integrates multimodality data for improved DTA prediction.
• To evaluate the performance of Modality-DTA against existing single-modality methods.

Main Methods:

• Developed Modality-DTA, a deep learning model leveraging multimodality drug and target data.
• Employed backward propagation neural networks for feature representation reconstruction.
• Utilized a tagging mechanism to reduce noise in latent representations from multimodality data.

Main Results:

• Modality-DTA significantly outperformed existing methods across three benchmark datasets.
• Achieved a 15.7% reduction in mean square error and a 12.74% improvement in area under the precision-recall curve on the Davis dataset.
• Identified Morgan fingerprint (drug) and one-hot encoding (target) as the most impactful modalities.

Conclusions:

• Modality-DTA represents a pioneering approach in exploring multimodality for DTA prediction.
• The findings highlight the importance of integrating diverse data sources for enhanced drug discovery predictions.
• Specific modalities like Morgan fingerprint and one-hot encoding demonstrate significant predictive power.