SCMFDD

SCMFDD predicts unobserved drug–disease associations by applying similarity-constrained matrix factorization that integrates drug feature-based similarities and disease semantic similarity to reveal latent drug and disease representations.

Key Features:

• Matrix Factorization with Biological Context: Employs similarity-constrained matrix factorization to project drug–disease association matrices into two low-rank spaces that capture latent features of drugs and diseases.
• Incorporation of Drug Features and Disease Semantics: Integrates drug feature-based similarities and disease semantic similarity as constraints on the low-rank spaces to guide biologically meaningful decompositions.
• High-Accuracy Performance: Demonstrates high predictive accuracy on benchmark datasets, evaluated using five-fold cross-validation and independent testing.
• Use of CTD Data: Utilizes known drug–disease associations from the Comparative Toxicogenomics Database (CTD) as input data.

Scientific Applications:

• Prediction of Novel Associations: Predicts unobserved drug–disease associations to support identification of novel therapeutic candidates.
• Target and Mechanism Discovery: Aids identification of therapeutic targets and interpretation of disease mechanisms through latent association patterns.
• Drug Development Prioritization: Supports prioritization of candidate treatments and reduces reliance on experimental screening by computationally ranking associations.

Methodology:

Uses known drug–disease associations (e.g., from CTD); applies similarity-constrained matrix factorization to project associations into two low-rank spaces; constrains the factorization with drug feature-based similarities and disease semantic similarity; evaluates performance via five-fold cross-validation and independent testing.