SSF-DDI: a deep learning method utilizing drug sequence and substructure features for drug-drug interaction...

Area of Science:

• Pharmacology
• Computational Chemistry
• Bioinformatics

Background:

• Drug-drug interactions (DDI) are common in combination therapy.
• Existing artificial intelligence methods for DDI prediction often neglect crucial sequence and substructure information of drug molecules.

Purpose of the Study:

• To develop a novel model for DDI prediction that incorporates both sequence and substructure features.
• To enhance the accuracy and comprehensiveness of DDI prediction by leveraging detailed molecular information.

Main Methods:

• Proposed the Sequence and Substructure Features for DDI (SSF-DDI) prediction model.
• Integrated drug sequence information with structural features derived from drug molecule graphs.

Main Results:

• The SSF-DDI model demonstrated superior performance compared to state-of-the-art DDI prediction models across various datasets.
• Achieved a 5.67% improvement in accuracy for predicting DDI involving unknown drugs.

Conclusions:

• The SSF-DDI model offers a more comprehensive and accurate approach to DDI prediction.
• The integration of sequence and substructure features is effective in improving DDI prediction, particularly for novel drug combinations.