Fig. 3
Quantitative performance on held-out experimental test data set of two different types of tasks for three high-quality gcPBM data sets. a The mean AUPRC for Mad, Max, and Myc plotted versus threshold at certain threshold percentiles of the data, when training with 2% of the data (left) and 10% of the data (right). In both cases 50 instances were randomly selected for training and performance of the 50 trained weights is evaluated on the same held-out test set. Error bars are the standard deviations. b Boxplot of Kendall’s τ on held-out test data set. Red ‘+’ indicate outliers, gray line represents the median. The bottom and top edges of the box represent the 25th and 75th percentiles, respectively
