Figure 2.

Protein-level predictive performance measured with AUC for the six partner-agnostic sequence-based predictors of RBRs on the benchmark dataset. This analysis focuses on the RNA-binding proteins as the calculation of the per-protein AUC is not possible for the other proteins. The violin plots in Panel (A) represent the distributions of the per-protein AUC values. The box plots inside the violin plots represent the first quartile (bottom of the box), the second quartile/median (white dot) and the third quartile (top of the box) for these distributions. The black points connected by the black solid lines denote the dataset-level AUC values. Panel (B) shows relation between per-protein AUC and the content of RBRs (fraction of RBR in the protein chain). The color-coded solid lines correspond to the linear fit between the content and the AUC values for a given predictor.