MSBMF

MSBMF integrates multiple biological similarity measures and applies non-negative bilinear matrix factorization with ADMM optimization to predict drug–disease associations for drug repositioning.

Key Features:

• Multi-Similarity Integration: Integrates similarity measures derived from target profiles, drug–drug interactions, side effects, and disease ontology.
• Separate Concatenation of Similarities: Concatenates separate similarity matrices for drugs and for diseases rather than fusing them into a single matrix to preserve distinct information within each measure.
• Matrix Factorization: Decomposes the drug–disease association matrix into two feature matrices representing drugs and diseases to extract latent features.
• Non-Negative Constraints: Applies non-negative matrix factorization constraints to maintain biological interpretability of the factorized matrices.
• Efficient Optimization: Uses an alternating direction method of multipliers (ADMM) algorithm to solve the optimization problem.

Scientific Applications:

• Drug Repositioning Prediction: Predicts new indications for existing or novel drugs by integrating multiple similarity measures.
• Therapeutic Candidate Identification: Identifies potential therapeutic uses for known drugs by ranking drug–disease associations.
• Improved Prediction Accuracy: Provides improved prediction accuracy compared to state-of-the-art methods.
• Support for Drug Discovery: Supports acceleration of drug discovery by prioritizing candidate drug–disease associations.

Methodology:

Calculate drug–drug and disease–disease similarities from multiple data sources; concatenate similarity matrices separately for drugs and diseases; apply bilinear matrix factorization to decompose the drug–disease association matrix into drug and disease feature matrices and extract latent features; enforce non-negative constraints during factorization; solve the resulting optimization problem using ADMM.