MOESM4 of In silico prediction of novel therapeutic targets using gene–disease association data

Additional file 4: Figure S4. Monte Carlo simulation to assess the effect of randomly sampling from the unlabelled class on the classifier performance. Ten thousands random samples of the unlabelled class were aggregated to the positive class and used to train and test a NN classifier. Histograms show distributions of (A) accuracy (mean = 0.71, standard deviation = 0.02) and (B) AUC (mean = 0.77, standard deviation = 0.02) calculated using the test set.