Detecting pore-lining regions in transmembrane protein sequences

BMC Bioinformatics

Table 1 SVM performance in predicting pore-lining residues, pore-lining helices and discriminating between monomeric and multimeric pore stoichiometry

Classifier	Precision	Recall	TPR	FPR	MCC	Accuracy
Pore-lining residue SVM	0.64	0.54	0.69	0.39	0.30	65%
Pore-lining residue (naïve)	0.53	0.23	0.23	0.21	0.03	51%
Pore-lining helix SVM	0.74	0.35	0.57	0.16	0.43	72%
Pore-lining helix (naïve)	0.50	0.17	0.22	0.17	0.15	55%
Monomeric/multimeric pore stoichiometry SVR	0.88	0.49	0.92	0.11	0.80	90%

TPR: True positive rate. FPR: False positive rate. Pore-lining helices were identified by averaging the raw SVM scores for each participating residue, while a threshold of 7 or more polar residues was found to be optimal for the naïve method.

ISSN: 1471-2105