CMLDR, a novel drug repositioning computational method based on Collaborative Metric Learning (CML)[1, 2], can recommend potential indications for known drugs having validated disease associations and new drugs without known associations.

1. Dataset.
(1) Drug_simMat.txt store drug similarity matrix;
(2) DrDiAssMat.txt stores known drug-disease association information;
2. Code.
(1) CMLDR-TN.py: predict potential indications for drugs;
(2) sampler.py : sample positive samples and negative samples.
(3) utils.py: split dataset into training, validation and test sets;
(4) evaluator.py: create evaluator for recall and precision evaluation;

[1] Hsieh, C. K. et al. (2017) Collaborative metric learning. In Proceedings of the 26th International Conference on World Wide Web, 193-201.

[2] Hsieh, C. K. Collaborative metric learning. https://github.com/changun/CollMetric, 2016.