MIF-DTI: a multimodal information fusion method for drug-target interaction prediction

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target interaction (DTI) prediction is crucial for developing new therapeutics and repurposing existing drugs.
• Current DTI prediction methods often suffer from limitations in single-source data encoding and insufficient multimodal information fusion.

Purpose of the Study:

• To develop an advanced DTI prediction method that effectively integrates multimodal information.
• To improve the accuracy and robustness of DTI prediction models.

Main Methods:

• Proposed a multimodal information fusion (MIF-DTI) method utilizing sequence and graph encoding modules to extract 1D sequence features and 2D topological structures from drugs and targets.
• Developed an ensemble version (MIF-DTI-B) by combining multiple MIF-DTI models via cross-validation for enhanced predictive performance.
• Employed a decoding module for effective fusion of diverse data modalities.

Main Results:

• Both MIF-DTI and MIF-DTI-B demonstrated superior performance compared to existing state-of-the-art methods on three public DTI datasets.
• The comprehensive integration of multimodal information was key to the improved predictive accuracy.

Conclusions:

• The proposed MIF-DTI and MIF-DTI-B models represent a significant advancement in DTI prediction by effectively leveraging multimodal data.
• These methods offer a promising approach for accelerating drug discovery and repurposing efforts.