HRWR

HRWR predicts potential efficacious drug combinations by performing random walk with restart on hypergraphs that encode high-order multi-drug interaction data to prioritize synergistic combinations for diseases such as cancer.

Key Features:

• High-Order Drug Combination Analysis: Incorporates multi-drug interaction data to analyze combinations beyond pairwise drug interactions.
• Hypergraph Representation: Models complex interactions among multiple drugs using hypergraphs to capture polypharmacy relationships.
• Random Walk with Restart Algorithm: Performs random walk with restart on hypergraphs, iteratively refining probability distributions of candidate combinations until convergence.
• Performance Evaluation: Validated using leave-one-out cross-validation (LOOCV) and assessed by Area Under the Receiver Operating Characteristic Curve (AUROC), yielding higher AUROC than comparator methods.

Scientific Applications:

• Lung cancer: Predicts efficacious multi-drug combinations for lung cancer by leveraging high-order interaction data.
• Breast cancer: Predicts efficacious multi-drug combinations for breast cancer by leveraging high-order interaction data.
• Colorectal cancer: Predicts efficacious multi-drug combinations for colorectal cancer by leveraging high-order interaction data.

Methodology:

Modeling drug combinations as hypergraphs and applying random walk with restart on these hypergraphs with iterative probability refinement until convergence, validated by leave-one-out cross-validation (LOOCV) and evaluated using AUROC.