On Classification of PDZ Domains: A Computational Study

Table 1

Comparisons between our methods and MODWT.

Methods	Sensitivity (%)			Specificity (%)			Positive predictive value (%)			Negative predictive value (%)
	WAD-1	WAD-2	MODWT	WAD-1	WAD-2	MODWT	WAD-1	WAD-2	MODWT	WAD-1	WAD-2	MODWT
Parametric classifier	81.82	91.67	81.82	66.67	100	55.56	75	100	69.23	75	90	71.43
K-nearest neighbors	72.73	75	63.64	33.33	77.78	88.89	57.14	81.82	87.5	50	70	66.67

Sensitivity gives true positive rate, or the recall rate of prediction algorithm. See (10).
Specificity gives true negative rate of prediction algorithm. See (11).
Positive predictive value confirms that a correct prediction is actually correct. See (12).
Negative predictive value confirms that a false prediction is actually false. See (13).
Parametric classifier and K-nearest neighbors are popular pattern classification algorithms. See [27].