T3SEdb: data warehousing of virulence effectors secreted by the bacterial Type III Secretion System

BMC Bioinformatics

Table 1 Performance measure of binary classifiers in WEKA for prediction of T3SEs.

	Training (10× cross-validation)			Testing
Binary classifier	Aroc	SE	SP	Aroc	SE	SP	PPV
Bayesian Logistic Regression	0.60	0.72	0.49	0.66	0.73	0.60	0.07
Support vector machines (SVM)	0.74	0.97	0.52	0.80	0.95	0.64	0.08
BayesNet	0.86	0.80	0.76	0.91	0.94	0.83	0.15
Naïve Bayes	0.89	0.84	0.82	0.93	0.91	0.83	0.17

SE, SP and PPV refer to sensitivity, specificity and positive predictive value measures, respectively. Aroc is the area under the receiver operator characteristic curve and is commonly used as a measure to assess the quality of a prediction model. The testing Aroc, SE and SP were done with a balanced dataset of 68 effectors and non-effectors that were not part of the training. The PPV was computed using an unbalanced dataset representing the approximate proportion of effectors and non-effectors in a bacterial genome.

ISSN: 1471-2105