UAMRL: multi-granularity uncertainty-aware multimodal representation learning for drug-target affinity prediction.

Area of Science:

• Computational chemistry
• Pharmacology
• Machine learning in drug discovery

Background:

• Computational prediction of drug-target affinity (DTA) is crucial for drug discovery.
• Deep learning models face challenges in interpretability and handling heterogeneous multimodal data.

Purpose of the Study:

• To develop a novel framework for reliable and interpretable DTA prediction.
• To address the limitations of existing deep learning models in handling multimodal data.

Main Methods:

• Proposed an Uncertainty-aware Multimodal Representation Learning (UAMRL) framework.
• Employed a dual-stream encoder for cross-modal association learning.
• Integrated an uncertainty quantification mechanism using the Normal-Inverse-Gamma distribution.

Main Results:

• UAMRL achieved superior predictive accuracy on multiple DTA datasets.
• The framework demonstrated improved prediction performance.
• Enhanced decision transparency in drug development applications.

Conclusions:

• UAMRL offers a reliable approach for DTA prediction by effectively integrating multimodal data and quantifying uncertainty.
• The framework improves both predictive accuracy and interpretability, aiding practical drug development.
• The proposed uncertainty quantification mechanism enhances the robustness of data fusion.