Integrative approaches to the prediction of protein functions based on the feature selection

BMC Bioinformatics

Table 5 Prediction quality of newly annotated genes

Specificity	# of genes	KLR	L1LR
		Exhaustive search	λ / λ _max = 0.01
			NOS*	S*
3-10	540	0.61	0.76	0.79
11-30	273	0.71	0.79	0.81
31-100	163	0.72	0.80	0.82
101-300	75	0.69	0.78	0.80

GO terms are categorized into four groups based on the number of genes covering the GO term. Prediction quality is estimated by using the AUC values for KLR with a data source selected by exhaustive search and KL1LR with all data sources. In the case of KL1LR, two different values of regularization parameter λ are used. NOS stands for non-standardization, and S for standardization.

ISSN: 1471-2105