CRYSTALP2: sequence-based protein crystallization propensity prediction

BMC Structural Biology

Table 2 Comparison of prediction quality measured via accuracy, MCC and AROC between the proposed and five competing methods.

Dataset	Method	Accuracy	MCC	AROC⁶
D418¹	SECRET	70.0	0.34	N/A
	CRYSTALP	77.5	0.55	N/A
	CRYSTALP2	77.5	0.55	N/A
TEST-RL²	CRYSTALP	46.5	-0.07	N/A
	SECRET	58.1	0.16	0.58
	ParCrys-W	67.4	0.38	0.84
	OB-Score	69.8	0.40	0.71
	ParCrys	79.1	0.58	0.84
	XtalPred⁴	76.7	0.54	0.82
	CRYSTALP2	69.8	0.40	0.72
TEST²	OB-Score	64.6	0.32	0.68
	ParCrys-W	68.0	0.37	0.75
	ParCrys	71.5	0.45	0.75
	XtalPred⁴	79.2	0.58	0.83
	CRYSTALP2	75.7	0.52	0.79
TEST-NEW	ParCrys³	70.6	0.43	0.75
	XtalPred⁴	70.0	0.40	0.76
	CRYSTALP2⁵	69.3	0.39	0.74

The AROC values for ParCrys, OB-Score and SECRET were taken from [23].
¹ Results based on tenfold cross-validation test on the D418 dataset
² Results based on training the classification model on FEAT dataset and testing on TEST-RL or TEST datasets, respectively
³ Results based the ParCrys server at http://www.compbio.dundee.ac.uk/xtal/cgi-bin/input.pl
⁴ Results based the XtalPred server at http://ffas.burnham.org/XtalPred-cgi/xtal.pl
⁵ Results based on training the classification model on FEAT dataset and testing on TEST-NEW datasets, respectively
⁶ N/A means that the corresponding results was not reported and cannot be duplicated or computed

ISSN: 1472-6807