FIGURE 2. Distributions of predicted and experimental enzyme activities.

A. Distribution of NAGLU relative enzyme activities 1) predicted for disease mutations in HGMD (HGMD, red); 2) predicted for inter-species variants (Interspecies, green); 3) predicted for mutations provided for the CAGI challenge (Prediction, blue), and 4) experimental activities for the challenge mutations (Experiment, purple). As expected, known disease mutations are predicted to have low activities and interspecies variant to have high activity. In contrast to these, the population variants have activities approximately equally distributed across the full range, for both prediction and experiment.

B. Relative yeast growth rate distributions for UBE2I (SUMO-ligase) mutation Set 1. The distribution of the unmapped predicted values (Submission 1, red) only approximately matches the experimental distribution (Experiment, black), available during the challenge. We submitted a second set of predictions in which each predicted value was mapped to the experimental value of closest rank (Submission 2, blue). This improves the overall match of the distributions (red and black) but not the prediction accuracy.