Drug-Drug Interaction Predictions via Knowledge Graph and Text Embedding: Instrument Validation Study.

Area of Science:

• Pharmacology
• Bioinformatics
• Computational Biology

Background:

• Drug-drug interactions (DDIs) pose significant clinical challenges, with pre-market detection being difficult.
• Existing big data approaches for DDI discovery are often noisy, incomplete, and focus only on binary interactions.

Purpose of the Study:

• To develop a novel framework for comprehensive DDI prediction.
• To address challenges of data noise, incompleteness, sparsity, and computational complexity in DDI prediction.

Main Methods:

• A framework, Predicting Rich DDI, was developed using graph embedding.
• Integrated large-scale drug knowledge graphs with biomedical text into a common low-dimensional space.
• Employed link prediction for efficient computation of rich DDI information.

Main Results:

• The proposed framework demonstrated superior capability and accuracy compared to state-of-the-art baseline methods.
• Extensive experiments on real-world datasets validated the model's effectiveness.

Conclusions:

• The Predicting Rich DDI framework enables multi-label DDI prediction across various domains.
• This is the first joint translation-based embedding model to integrate drug knowledge graphs and biomedical text for DDI learning.
• The model efficiently predicts multiple DDI labels, outperforming existing baselines.