A Hypergraph-Based Model for Predicting Potential Drug Combinations in Cancer Therapy.

Area of Science:

• Computational biology
• Drug discovery
• Network science

Background:

• Drug combinations enhance efficacy and overcome resistance in complex diseases.
• Existing computational models often fail to capture higher-order drug synergies beyond pairwise interactions.

Purpose of the Study:

• To develop an advanced computational model for predicting effective multi-drug combinations.
• To address the limitations of pairwise interaction prediction in current drug discovery models.

Main Methods:

• Introduced an enhanced hypergraph random walk (EHRW) model to represent multi-drug relationships.
• Incorporated a post-processing strategy using chemical similarity (SMILES fingerprints) to refine predictions.
• Validated the model on lung and breast cancer datasets.

Main Results:

• The enhanced EHRW model demonstrated superior predictive accuracy compared to baseline models.
• Chemical similarity-based post-processing significantly improved prediction performance.
• The model successfully extended to scoring three-drug combinations.

Conclusions:

• The enhanced EHRW model provides a flexible, accurate, and scalable tool for discovering effective multi-drug regimens.
• Integrating network topology with chemical similarity enhances drug combination prediction.
• This approach facilitates the evaluation of higher-order drug therapies for complex diseases.