TMB-Hunt: An amino acid composition based method to screen proteomes for beta-barrel transmembrane proteins

BMC Bioinformatics

Table 4 Comparison of various predictors with specific examples.

	BOMP	Prof-TMB	Pred-TMBB	TMB-Hunt: Leave Homs Out	TMB-Hunt: Leave One Out
NalP – Q8GKS5	0^Γ	12.32	2.92	10.73	10.73
TSX – P22786	1	10.92	2.94	4.47	4.47
FadL – P10384	1	9.47	2.88	0.8	0.8
BtuB – P06129	1	10.39	2.91	10.82	10.82
Secretin – P31700	0^Γ	3.73^Γ	2.90	5.48	5.48
Usher – P30130	1	10.46	2.95	10.79	10.79
60 kDA cysteine rich OMP – P26758	0^Γ	2.42^Γ	3.03^Γ	-1.70^Γ	-1.70^Γ
Mycobacterial Porin – Q9RLP7	0^Γ	5.65^Γ	2.84	7.74	7.74
TolC – P02930	0^Γ	1.85^Γ	2.90	6.76	10.64
Alpha hemolysin – O68404	0^Γ	0.83^Γ	2.88	9.89	9.89
VDAC – Q60931	1	6.55	2.88	5.24	5.24
Tom40 – Q18090	1	4.79^Γ	2.92	-1.04^Γ	-1.04^Γ
Toc75 – Q43715	0^Γ	6.50	2.99^Γ	-1.41^Γ	1.24
OEP24 – O49929	0^Γ	3.11^Γ	2.87	1.55	1.55

All programs were run via their web interfaces, using default settings. Sequences classified as non-bbtm are marked using ^Γ. BOMP [22] values indicate the number bbtm proteins predicted given the number of sequences queried. Prof-TMB [17] returns a z-score statistic for which 50% of bbtm proteins get a z-score of >= 10 at an accuracy of 80% and 35% bbtm proteins get a z-score >= 6 at an accuracy of 35%. Pred-TMBB [16] returns a threshold score, for which sequences with threshold scores >2.965 are assumed not to be bbtm proteins. Beta-barrel scores, were given for TMB-Hunt. These were calculated without inclusion of evolutionary information, using 'leave homologues out' and 'leave one out' cross-validations. Beta-barrel scores >0 indicate that there is a greater probability that the sequence is from a bbtm protein.

ISSN: 1471-2105