. 2019 Jul 5;35(14):i343–i353. doi: 10.1093/bioinformatics/btz324

Table 2.

Comparison of the predictive performance of SCRIBER with the seven representative predictors of PBRs on the TEST dataset

Predictor		Sensitivity	Specificity	Precision	Accuracy	F1	MCC	AUC	AUPRC	AULC	AULCratio	AUCPC	CPR
SPPIDER	Average	0.202	0.870	0.194	0.781	0.198	0.071	0.517	0.159	0.015	1.791	0.596	0.332
	Stdev	±0.011	±0.003	±0.010	±0.004	±0.010	±0.009	±0.005	±0.007	±0.001	±0.099	±0.007	±0.013
	P-value	*5.40E-15*	*2.45E-11*	*1.81E-15*	*1.19E-11*	*2.84E-15*	*6.81E-16*	*1.24E-19*	*3.82E-16*	*2.55E-08*	*4.26E-14*	*4.86E-22*	*1.85E-19*
SPRINT	Average	0.183	0.873	0.183	0.781	0.183	0.057	0.570	0.167	0.012	1.523	0.663	0.380
	Stdev	±0.010	±0.004	±0.009	±0.005	±0.010	±0.008	±0.008	±0.008	±0.001	±0.096	0.006	±0.012
	P-value	*2.90E-16*	*3.02E-10*	*3.13E-16*	*2.61E-11*	*2.93E-16*	*1.19E-16*	*1.42E-16*	*1.71E-15*	*2.29E-12*	*4.15E-15*	*9.27E-24*	*2.41E-21*
PSIVER	Sverage	0.191	0.874	0.191	0.783	0.191	0.066	0.581	0.170	0.013	1.607	0.537	0.250
	Stdev	±0.012	±0.003	±0.012	±0.005	±0.012	±0.012	±0.010	±0.009	±0.001	±0.118	±0.011	±0.010
	P-value	*3.70E-15*	*5.02E-10*	*4.51E-15*	*6.38E-11*	*3.96E-15*	*1.82E-15*	*2.78E-15*	*6.96E-15*	*9.74E-12*	*1.38E-14*	*1.62E-19*	*5.44E-17*
SPRINGS	Average	0.229	0.882	0.228	0.796	0.229	0.111	0.625	0.201	0.015	2.164	0.500	0.236
	Stdev	±0.013	±0.003	±0.012	±0.004	±0.012	±0.011	±0.007	±0.011	±0.001	±0.118	±0.007	±0.008
	P-value	*1.14E-12*	*1.75E-07*	*1.01E-12*	*1.69E-08*	*1.05E-12*	*4.41E-13*	*3.99E-13*	*4.02E-12*	*1.38E-08*	*2.62E-12*	*2.54E-19*	*4.44E-17*
LORIS	Average	0.264	0.887	0.263	0.805	0.263	0.151	0.656	0.228	0.017	2.726	0.439	0.195
	Stdev	±0.010	±0.003	±0.010	±0.004	±0.010	±0.009	±0.006	±0.010	±0.001	±0.101	±0.004	±0.003
	P-value	*1.90E-10*	*5.77E-05*	*2.19E-10*	*9.53E-06*	*2.00E-10*	*1.35E-10*	*2.25E-10*	*9.14E-10*	*1.66E-05*	*2.73E-09*	*3.48E-17*	*1.27E-15*
CRFPPI	Average	0.268	0.887	0.264	0.805	0.266	0.154	0.681	0.238	0.017	2.671	0.448	0.208
	Stdev	±0.011	±0.003	±0.010	±0.004	±0.010	±0.009	±0.006	±0.012	±0.001	±0.099	±0.007	±0.009
	P-value	*8.36E-10*	*5.41E-05*	*2.54E-10*	*1.16E-05*	*4.28E-10*	*2.17E-10*	*1.64E-06*	*6.22E-08*	*4.41E-05*	*1.11E-09*	*3.55E-17*	*1.98E-14*
SSWRF	Average	0.288	0.891	0.286	0.811	0.287	0.178	0.687	0.256	0.018	3.093	0.423	0.209
	Stdev	±0.010	±0.002	±0.010	±0.003	±0.010	±0.009	±0.006	±0.013	±0.001	±0.121	±0.006	±0.005
	P-value	*1.30E-07*	*3.83E-03*	*1.25E-07*	*1.15E-03*	*1.22E-07*	*8.98E-08*	*1.52E-05*	*4.89E-05*	*7.27E-03*	*3.14E-06*	*4.81E-16*	*3.70E-16*
SCRIBER	Average	0.334	0.896	0.332	0.821	0.333	0.230	0.715	0.287	0.020	3.725	0.282	0.116
SCRIBER	Stdev	±0.013	±0.004	±0.013	±0.007	±0.013	±0.016	±0.013	±0.012	±0.001	±0.258	±0.014	±0.009

Note: The methods are sorted by their AUC value in the ascending order. The binary predictions for all methods are calibrated to allow for direct side-by-side comparison, such that the number of putative PBRs each method generates equals to the number of the native PBRs. The results are computed over 10 subsets of randomly selected 50% of TEST proteins to assess robustness of the predictions and evaluate statistical significance of differences between SCRIBER and the other methods (Section 2.3 gives details). We report the corresponding averages, standard deviations (stdev) and P-values; differences with P-value <0.05 are assumed statistically significant and are given in bold italics font. The best value for each measure of predictive quality are shown with bold font.