MLMC

MLMC predicts novel drug-disease associations by integrating multi-view learning and matrix completion to support drug repositioning.

Key Features:

• Multi-View Learning (ML): Constructs comprehensive similarity matrices by combining five distinct drug similarity metrics and two disease similarity matrices using Laplacian graph regularization to capture complex relationships between drugs and diseases.
• Matrix Completion (MC): Employs matrix completion under a low-rank structure assumption to introduce additional positive entries and enhance the initial drug-disease association matrix.
• Iterative Learning and Integration: Re-executes the multi-view learning algorithm after matrix completion and synthesizes results from the initial and MC-enhanced iterations to produce final drug-disease association predictions.

Scientific Applications:

• Drug Repositioning: Identifies novel therapeutic uses for existing drugs by predicting potential drug-disease associations, validated using 10-fold cross-validation and de novo tests against state-of-the-art approaches.

Methodology:

Initial multi-view learning constructs similarity matrices from multiple drug and disease data sources using Laplacian graph regularization and updates the initial drug-disease association matrix; matrix completion then introduces additional positive associations under a low-rank assumption followed by re-execution of the multi-view learning algorithm and integration of both results.