BNNR

BNNR (Bounded Nuclear Norm Regularization) is a computational method for drug repositioning to identify new indications for existing drugs. The method models drug repositioning as a recommendation system problem and employs matrix completion techniques to predict novel drug-disease associations.

Key features of BNNR:

1. Assumes that the underlying latent factors determining drug-disease associations are highly correlated, resulting in a low-rank drug-disease matrix.

2. Tolerates noisy drug-drug and disease-disease similarities by incorporating a regularization term to balance the approximation error and rank properties.

3. Incorporates additional constraints to ensure all predicted matrix entry values are within a specific interval.

4. Operates on an adjacency matrix of a heterogeneous drug-disease network, integrating drug-drug, drug-disease, and disease-disease networks.

5. Capable of dealing with cold start problems naturally.