FSSM-DDI: Fusion State Space Model for predicting drug-drug interaction using social-media and drug descriptions

Area of Science:

• Computational chemistry
• Bioinformatics
• Artificial intelligence in medicine

Background:

• Predicting drug-drug interactions (DDIs) is vital for safe medication use and pharmaceutical research.
• Current deep learning models face challenges with long-range dependencies and high computational costs.
• This limits their use in large-scale drug safety assessments.

Purpose of the Study:

• To develop a computationally efficient model for drug-drug interaction prediction (DDIP).
• To address limitations of existing models in handling long-range dependencies and multimodal data.

Main Methods:

• Proposed a Fusion State Space Model (FSSM) incorporating a selective State Space Model (SSM).
• Integrated an Interaction-based Selective Filtering (ISF) module to manage multimodal information redundancy.
• Evaluated the model on the DDIExtraction-2013 corpus.

Main Results:

• FSSM achieved an F1-score of 75.23%, comparable to state-of-the-art Transformer models.
• FSSM demonstrated significantly reduced training time compared to existing methods.
• Ablation studies confirmed the contribution of social-media embedding and the ISF module.

Conclusions:

• FSSM provides an effective balance between predictive accuracy and computational efficiency for DDIP.
• The model's linear complexity facilitates large-scale screening and real-time applications.
• FSSM is a valuable tool for drug discovery and enhancing drug safety.