A multi-layer hypergraph framework for drug-drug interaction prediction based on transformer and hypergraph...

Area of Science:

• Pharmacology and Cheminformatics
• Artificial Intelligence in Drug Discovery

Background:

• Drug-drug interactions (DDIs) pose significant challenges in drug research and clinical practice.
• Existing network models struggle to capture complex, multi-element synergistic drug interactions.
• Accurate DDI prediction is crucial for enhancing treatment safety and optimizing medication regimens.

Purpose of the Study:

• To develop an advanced framework for predicting drug-drug interactions (DDIs) that overcomes limitations of conventional models.
• To effectively represent and analyze multi-element synergistic interactions among drugs.
• To improve the accuracy and reliability of DDI prediction.

Main Methods:

• Proposed a Multi-Layer Hypergraph framework for drug interaction prediction using Transformer and Hypergraph Convolution (MLHTHC).
• Constructed a multi-layer drug similarity hypergraph using chemical structure, ATC code, drug category, and target information.
• Employed spectral Hamming similarity for layer weight determination, hypergraph convolutional networks for node embedding, Transformer for feature fusion, and MLP for DDI prediction.

Main Results:

• The MLHTHC model demonstrated superior performance compared to existing methods like DPSP and DANN.
• The integration of Transformer and hypergraph convolution significantly enhanced prediction accuracy for drug-drug interactions.
• The framework effectively captures complex, multi-element synergistic drug relationships.

Conclusions:

• The proposed MLHTHC framework offers a powerful and effective tool for predicting drug-drug interactions.
• This approach advances the field of computational pharmacology by improving DDI prediction capabilities.
• Accurate DDI prediction using MLHTHC can lead to safer and more effective pharmacotherapy.