Fig. 2

PRODE robustly identifies essential genes and biological processes across different screening datasets. a Diagram shows the collection of genes employed for the establishment of the reference essential genes set. b Barplots demonstrate the superior or similar performance—assessed through ROC AUC (left) and PR AUC (right) of PRODE compared to other approaches when classifying reference essential and non-essential gene lists using sgRNA (left) and shRNA (right) dependency scores. AUCs are reported along with 95th confidence interval (also in Additional file 1: Fig. S1a–b). c PRODE scores exhibit greater consistency (Spearman’s correlation coefficient on the y-axis) compared to average gene effects across all cell lines when subjected to increasing levels Gaussian of noise in the input dependency scores (x-axis). d PRODE demonstrates a higher correlation (Spearman’s correlation of 0.58) between sgRNA and shRNA-derived essentiality scores compared to alternative methods. e PRODE essentiality scores exhibit a higher Spearman’s correlation with coefficients of variation of gene expression across cell lines (d; 0.62 and 0.43, respectively, for sgRNA and shRNA dependency scores). Note: a positive correlation emerges when low NIE scores correspond to low coefficient of variation. f PRODE displays a higher correlation with genes conservation scores. For the ease of interpretation, a positive correlation results when highly conserved genes across species display lower NIE scores