Evaluation of methods for predicting the topology of β-barrel outer membrane proteins and a consensus prediction method

Table 1 Obtained accuracy of the predictors, in a test set of 20 outer membrane proteins

METHOD	TYPE	SEQUENCE	Q_β	C_β	SOV	Correctly predicted topologies	Correctly predicted barrel size
HMM-B2TMR	HMM	barrel	0.737	0.557	0.836	15	17
		precursor	0.790	0.600	0.813	14	16
ProfTMB	HMM	barrel	0.734	0.537	0.818	14	17
		precursor	0.777	0.575	0.784	12	16
PRED-TMBBpost	HMM	barrel	0.818	0.630	0.886	14	19
		precursor	0.842	0.637	0.852	14	16
PRED-TMBBnbest	HMM	barrel	0.818	0.629	0.877	12	17
		precursor	0.849	0.637	0.856	11	13
TBBPred-comb	NN+SVM	barrel	0.702	0.428	0.664	0	0
		precursor	0.701	0.424	0.496	0	0
TBBPred-nn	NN	barrel	0.735	0.466	0.672	0	1
		precursor	0.726	0.432	0.496	0	1
TBBPred-svm	SVM	barrel	0.744	0.458	0.721	1	3
		precursor	0.744	0.426	0.535	0	0
B2TMPRED	NN	barrel	0.723	0.498	0.738	7	9
		precursor	0.709	0.466	0.551	0	0
TMBETA-NET	HMM	barrel	0.697	0.415	0.698	3	8
		precursor	0.663	0.353	0.515	0	4
BETA-TM	NN	barrel	0.690	0.395	0.691	1	2
		precursor	0.663	0.322	0.497	0	1
PSI-PRED	NN	barrel	0.731	0.484	0.690	0	0
		precursor	0.756	0.495	0.569	0	0
HMM-B2TMR, ProfTMB, PRED-TMBBpost, B2TMPRED, TBBPred-nn	CONSENSUS	barrel	0.819	0.641	0.924	18	20
		precursor	0.849	0.660	0.874	15	18

For an explanation of the measures of accuracy see the Materials and Methods section. Abbreviations: PRED-TMBBpost: PRED-TMBB method with posterior decoding, PRED-TMBBnbest: PRED-TMBB method with NBest decoding, TBBPred-nn: The Neural Network module of TBBPred, TBBPred-svm: The SVM module of TBBPred, TBBPred-comb: TBBPred, combining the Neural Network and SVM modules. The performance of the best individual predictor, and the best available consensus obtained are highlighted with bold.

ISSN: 1471-2105