Learning from positive examples when the negative class is undetermined- microRNA gene identification

Table 1 One-class results obtained from the secondary features plus sequence features.

	C. elegans			Mouse			Human			All-miRNA
Method	Sen	Spe	MCC	Sen	Spe	MCC	Sen	Spe	MCC	Sen	Spe	MCC	Average MCC
OC-SVM	0.73	0.93	0.67	0.80	0.93	0.74	0.72	0.99	0.74	0.69	0.91	0.62	0.70
OC-Gaussian	0.84	0.93	0.77	0.89	0.93	0.82	0.82	0.99	0.82	0.82	0.99	0.82	0.81
OC-Kmeans	0.79	0.93	0.73	0.85	0.92	0.77	0.89	0.92	0.81	0.89	0.80	0.69	0.75
OC-PCA	0.87	0.89	0.76	0.88	0.92	0.80	0.90	0.79	0.69	0.90	0.86	0.76	0.77
OC-KNN	0.90	0.86	0.76	0.90	0.92	0.82	0.90	0.96	0.86	0.90	0.93	0.83	0.82
Two-Class
Naïve Bayes	0.89	0.93	0.82 (125)	0.93	0.97	0.90 (200)	0.99	0.92	0.92 (300)	0.97	0.96	0.93 (4000)	0.88
SVM	0.90	0.97	0.87 (200)	0.95	0.98	0.93 (500)	0.99	0.99	0.98 (300)	0.98	0.95	0.93 (900)	0.92

Sen = sensitivity, Spe = specificity, and MCC = Matthews Correlation Coefficient. Results are presented for four genomes individually (C. elegans, Mouse, and Human) and All-miRNA as a mixture of multiple miRNAs species. The number in parentheses is the corresponding number of optimal negative examples giving the highest MCC.

ISSN: 1748-7188