Fig. 1.

Framework for prediction of therapeutic targets based on disease association data. Observations and features were gathered from Open Targets while the labels were collected from Informa Pharmaprojects. The resulting data matrix was split into a prediction set and a working dataset. The former was kept aside and used to perform predictions once the model was established. The latter, containing both positive and unlabelled observations, was further split into training and test sets to train the classifiers and evaluate their performance