Fig. 8

ESCOTT and PRESCOTT benign versus pathogenic mutations. A Precision and Recall curves for the 500 human proteins used to calibrate the two thresholds in ESCOTT and PRESCOTT for the purpose of distinguishing between benign and pathogenic variants. B Same as Fig. 2E where a colored background indicates those mutations that are labeled as pathogenic (red), benign (blue), VUS (white) variants by ESCOTT/PRESCOTT. C The 48 mutations considered in B are colored with ClinVar classification as benign/likely benign (16, blue) and pathogenic/likely pathogenic (32, red). Only 45 points are visible, as described in the legend of Fig. 2E. D The set of ACMG benign mutations. The plot on the left reports the exact count of mutations which are wrongly predicted as pathogenic (left), or VUS (centre) and correctly as benign (right) by each method. The plot on the right reports, for each protein in the dataset, the percentage of correctly predicted mutations. The horizontal lines indicate the average over all percentages obtained by each method. E The set of ACMG pathogenic mutations is analyzed as in D. F The set of gain-of-function pathogenic mutations involved in autoinflammatory diseases is analyzed as in D. For each gene, the corresponding UniProt ID is provided in parentheses