MQAPRank: improved global protein model quality assessment by learning-to-rank

BMC Bioinformatics

Table 3 The performances of the MQAPRank and several leading methods on CASP11 dataset based on GDT_TS score

Method	Method Type	Best 150						Sel20
Method	Method Type	Diff	MCC	AUC	Loss	mPCC^a	PCC^b	Diff	MCC	AUC	Loss	mPCC	PCC
MQAPRank	quasi-clustering	5.78	0.87	0.98	4.32	0.74	0.95	6.47	0.78	0.97	9.55	0.77	0.91
MULTICOM-REFINE	clustering	6.06	0.87	0.98	7.62	0.68	0.94	7.99	0.61	0.98	5.20	0.90	0.92
DAVIS-QAconsensus	clustering	6.17	0.87	0.98	7.74	0.68	0.94	7.33	0.62	0.98	5.51	0.90	0.95
Pcons-net	clustering	7.50	0.81	0.98	5.28	0.71	0.94	9.08	0.57	0.98	2.79	0.91	0.93
MULTICOM-CLUSTER	single	13.2	0.66	0.91	7.06	0.43	0.79	12.4	0.62	0.92	9.47	0.71	0.82
MQAPsingleA	quasi-single	13.8	0.60	0.90	8.95	0.65	0.75	9.66	0.68	0.95	3.64	0.92	0.88

^amPCC: mean Pearson’s correlation coefficient between the predicted and GDT_TS scores of per target protein
^bPCC: Pearson’s correlation coefficient between the predicted and GDT_TS scores on overall models. Bold value indicates highest performance on corresponding evaluation metric

ISSN: 1471-2105