Table 5.

Predictive performance of the eight selected sequence-based predictors of RBRs, HybridRNAbind and the structure-based PST-PRNA on the version of the test dataset that excludes proteins similar to the training dataset of PST-PRNA and the two subsets of this dataset that include the structure-annotated and the disorder-annotated test proteins. The predictor with the highest AUC for each protein set is identified with the bold font. Results of the statistical significance test are shown next to the reported values using the ‘x/y’ format where x represent comparison against the current method with the best AUC (i.e. PST-PRNA for the structure-annotated proteins, DeepDISOBind for the disorder-annotated proteins, and NCBRPred for the complete test set) and y represent comparison against HybridRNAbind; +/ = /– means that the best current predictor or HybridRNAbind is significantly better/not different/significantly worse than another method at P-value < 0.01

		Structure-annotated proteins		Disorder-annotated proteins		All proteins
Predictor type	Predictor	AUC	AULCratio	AUC	AULCratio	AUC	AULCratio
Structure-trained predictors	Pprint	0.775+/+	8.951+/+	0.604+/+	3.411+/+	0.632+/+	4.280+/+
	BindN+	0.455+/+	0.012+/+	0.493+/+	0.019+/+	0.487+/+	0.017+/+
	DRNApred	0.494+/+	0.001+/+	0.614+/+	2.832+/+	0.592+/+	2.574+/+
	NCBRPred	0.847 = /=	32.218−/=	0.656+/+	2.734+/+	0.684 /+	6.432 /+
	ProNA2020	0.598+/+	12.509−/+	0.413+/+	1.129+/+	0.440+/+	1.895+/+
	MTDsite	0.787+/+	10.017 = /+	0.576+/+	2.833+/+	0.609+/+	3.898+/+
Disorder-trained predictors	DisoRDPbind	0.664+/+	3.323+/+	0.520+/+	2.980+/+	0.536+/+	3.061+/+
	DeepDISObind	0.566+/+	0.723+/+	0.704 /+	5.140 /=	0.683 = /+	4.927+/+
Structure-based predictor	PST-PRNA	0.851 /−	11.165 /+	0.613+/+	3.534+/+	0.658+/+	3.937+/+
Meta predictor	HybridRNAbind	0.832+/	32.351−/	0.713−/	4.856 = /	0.738−/	8.097−/