Fig. 9
From: Target prediction utilising negative bioactivity data covering large chemical space
Flow of training and query data through the prediction approach described in this work. 1. The PubChem inactives and ChEMBL actives used to train the Bernoulli Naïve Bayes classifier. 2. The query data is imported and classification is performed producing probability scores. 3. If binary predictions are required, class-specific cut-off thresholds can be applied. F, A, P and R represent the F1, accuracy, precision and recall metrics used to calculate thresholds. Block arrows represent the flow of training data, while dashed arrows represent query fingerprint or prediction score data flow