Table 1. Open in new tab Comparison of...

Predictor	AUC	Sensitivity	FPR (sensitivity to FPR rate)	FCPR (sensitivity to FCPR rate)
CRFPPI	0.683	0.706	0.271	0.261^*	0.113 [2.4]	0.097= [2.7]	0.204 [1.3]	0.182^* [1.4]
SSWRF	0.693	0.701=	0.311	0.295^*	0.113 [2.8]	0.105= [2.8]	0.210 [1.5]	0.191^* [1.5]
LORIS	0.657	0.671^*	0.266	0.260^*	0.114 [2.3]	0.109= [2.4]	0.192 [1.4]	0.167^* [1.6]
SPRINGS	0.626	0.646^*	0.234	0.219^*	0.120 [2.0]	0.103= [2.1]	0.235 [1.0]	0.212^* [1.0]
SCRIBER	0.717	0.635^*	0.311	0.192^*	0.093 [3.3]	0.046 [4.2]	0.100 [3.1]	0.000 [inf]
SPRINT	0.573	0.608^*	0.187	0.156^*	0.128 [1.5]	0.110= [1.4]	0.379 [0.5]	0.409^* [0.4]
PSIVER	0.578	0.606^*	0.192	0.157^*	0.128 [1.5]	0.108= [1.5]	0.251 [0.8]	0.200^* [0.8]
DeepPPISP	0.642	0.599^*	0.477	0.500	0.286 [1.7]	0.360^* [1.4]	0.422 [1.1]	0.500^* [1.0]
SPPIDER	0.513	0.486^*	0.198	0.125^*	0.132 [1.5]	0.102= [1.2]	0.323 [0.6]	0.293^* [0.4]

Predictor

AUC

Sensitivity

FPR (sensitivity to FPR rate)

FCPR (sensitivity to FCPR rate)

Dataset-level

Median per-protein

Dataset-level

Median per-protein

Dataset-level

Median per-protein

Dataset-level

Median per-protein

CRFPPI

0.683

0.706

0.271

0.261^*

0.113 [2.4]

0.097= [2.7]

0.204 [1.3]

0.182^* [1.4]

SSWRF

0.693

0.701=

0.311

0.295^*

0.113 [2.8]

0.105= [2.8]

0.210 [1.5]

0.191^* [1.5]

LORIS

0.657

0.671^*

0.266

0.260^*

0.114 [2.3]

0.109= [2.4]

0.192 [1.4]

0.167^* [1.6]

SPRINGS

0.626

0.646^*

0.234

0.219^*

0.120 [2.0]

0.103= [2.1]

0.235 [1.0]

0.212^* [1.0]

SCRIBER

0.717

0.635^*

0.311

0.192^*

0.093 [3.3]

0.046 [4.2]

0.100 [3.1]

0.000 [inf]

SPRINT

0.573

0.608^*

0.187

0.156^*

0.128 [1.5]

0.110= [1.4]

0.379 [0.5]

0.409^* [0.4]

PSIVER

0.578

0.606^*

0.192

0.157^*

0.128 [1.5]

0.108= [1.5]

0.251 [0.8]

0.200^* [0.8]

DeepPPISP

0.642

0.599^*

0.477

0.500

0.286 [1.7]

0.360^* [1.4]

0.422 [1.1]

0.500^* [1.0]

SPPIDER

0.513

0.486^*

0.198

0.125^*

0.132 [1.5]

0.102= [1.2]

0.323 [0.6]

0.293^* [0.4]

Predictor	AUC	Sensitivity	FPR (sensitivity to FPR rate)	FCPR (sensitivity to FCPR rate)
CRFPPI	0.683	0.706	0.271	0.261^*	0.113 [2.4]	0.097= [2.7]	0.204 [1.3]	0.182^* [1.4]
SSWRF	0.693	0.701=	0.311	0.295^*	0.113 [2.8]	0.105= [2.8]	0.210 [1.5]	0.191^* [1.5]
LORIS	0.657	0.671^*	0.266	0.260^*	0.114 [2.3]	0.109= [2.4]	0.192 [1.4]	0.167^* [1.6]
SPRINGS	0.626	0.646^*	0.234	0.219^*	0.120 [2.0]	0.103= [2.1]	0.235 [1.0]	0.212^* [1.0]
SCRIBER	0.717	0.635^*	0.311	0.192^*	0.093 [3.3]	0.046 [4.2]	0.100 [3.1]	0.000 [inf]
SPRINT	0.573	0.608^*	0.187	0.156^*	0.128 [1.5]	0.110= [1.4]	0.379 [0.5]	0.409^* [0.4]
PSIVER	0.578	0.606^*	0.192	0.157^*	0.128 [1.5]	0.108= [1.5]	0.251 [0.8]	0.200^* [0.8]
DeepPPISP	0.642	0.599^*	0.477	0.500	0.286 [1.7]	0.360^* [1.4]	0.422 [1.1]	0.500^* [1.0]
SPPIDER	0.513	0.486^*	0.198	0.125^*	0.132 [1.5]	0.102= [1.2]	0.323 [0.6]	0.293^* [0.4]

Table 1.

Open in new tab

Comparison of the predictive performance of nine representative predictors of PBRs on the benchmark dataset

Predictor	AUC		Sensitivity		FPR (sensitivity to FPR rate)		FCPR (sensitivity to FCPR rate)
Predictor	Dataset-level	Median per-protein	Dataset-level	Median per-protein	Dataset-level	Median per-protein	Dataset-level	Median per-protein
CRFPPI	0.683	0.706	0.271	0.261^*	0.113 [2.4]	0.097= [2.7]	0.204 [1.3]	0.182^* [1.4]
SSWRF	0.693	0.701=	0.311	0.295^*	0.113 [2.8]	0.105= [2.8]	0.210 [1.5]	0.191^* [1.5]
LORIS	0.657	0.671^*	0.266	0.260^*	0.114 [2.3]	0.109= [2.4]	0.192 [1.4]	0.167^* [1.6]
SPRINGS	0.626	0.646^*	0.234	0.219^*	0.120 [2.0]	0.103= [2.1]	0.235 [1.0]	0.212^* [1.0]
SCRIBER	0.717	0.635^*	0.311	0.192^*	0.093 [3.3]	0.046 [4.2]	0.100 [3.1]	0.000 [inf]
SPRINT	0.573	0.608^*	0.187	0.156^*	0.128 [1.5]	0.110= [1.4]	0.379 [0.5]	0.409^* [0.4]
PSIVER	0.578	0.606^*	0.192	0.157^*	0.128 [1.5]	0.108= [1.5]	0.251 [0.8]	0.200^* [0.8]
DeepPPISP	0.642	0.599^*	0.477	0.500	0.286 [1.7]	0.360^* [1.4]	0.422 [1.1]	0.500^* [1.0]
SPPIDER	0.513	0.486^*	0.198	0.125^*	0.132 [1.5]	0.102= [1.2]	0.323 [0.6]	0.293^* [0.4]

Predictor	AUC		Sensitivity		FPR (sensitivity to FPR rate)		FCPR (sensitivity to FCPR rate)
Predictor	Dataset-level	Median per-protein	Dataset-level	Median per-protein	Dataset-level	Median per-protein	Dataset-level	Median per-protein
CRFPPI	0.683	0.706	0.271	0.261^*	0.113 [2.4]	0.097= [2.7]	0.204 [1.3]	0.182^* [1.4]
SSWRF	0.693	0.701=	0.311	0.295^*	0.113 [2.8]	0.105= [2.8]	0.210 [1.5]	0.191^* [1.5]
LORIS	0.657	0.671^*	0.266	0.260^*	0.114 [2.3]	0.109= [2.4]	0.192 [1.4]	0.167^* [1.6]
SPRINGS	0.626	0.646^*	0.234	0.219^*	0.120 [2.0]	0.103= [2.1]	0.235 [1.0]	0.212^* [1.0]
SCRIBER	0.717	0.635^*	0.311	0.192^*	0.093 [3.3]	0.046 [4.2]	0.100 [3.1]	0.000 [inf]
SPRINT	0.573	0.608^*	0.187	0.156^*	0.128 [1.5]	0.110= [1.4]	0.379 [0.5]	0.409^* [0.4]
PSIVER	0.578	0.606^*	0.192	0.157^*	0.128 [1.5]	0.108= [1.5]	0.251 [0.8]	0.200^* [0.8]
DeepPPISP	0.642	0.599^*	0.477	0.500	0.286 [1.7]	0.360^* [1.4]	0.422 [1.1]	0.500^* [1.0]
SPPIDER	0.513	0.486^*	0.198	0.125^*	0.132 [1.5]	0.102= [1.2]	0.323 [0.6]	0.293^* [0.4]

Note: The methods are sorted by their median per-protein AUC values in the descending order. Methods indicated in bold font provide the best value of a given measure of predictive performance. We report medians of the per-protein values and assess significance of the differences between the per-protein values of the best method and each of the other methods.

Statistically significant differences (P-value <0.001), while ‘=’ denotes differences that are not significant (P-value ≥ 0.001). We use paired t-test (for normal data) or Wilcoxon test (otherwise) and we assess normality with the Kolmogorov−Smirnov test.

This Feature Is Available To Subscribers Only