SCPRED: Accurate prediction of protein structural class for sequences of twilight-zone similarity with predicting sequences

Table 7 Summary of the feature selection results.

Feature set		# features
		all	after step 1	after step 2
Length		1	0	0
Index-based		50	5	0
CV and CMV		60	2	0
CV for collocated AAs		2000	4	1
Property group-based		35	1	0
Predicted secondary structure content		4	2	0
Predicted secondary structure-based	with PSI-PRED	86	27	8
	with YASPIN	86	12	0
Total # of features		2322	53	9
10 fold cross validation accuracy for prediction on 25PDB dataset		73.2%	80.2%	80.1%

The "feature set" columns defines categories of the considered features, the "all" column shows the total number of features in a given category, while the "after step 1" and "after step 2" columns show the corresponding number of features from a given category that were selected in the step 1 and step 2 of the feature selection procedure, respectively.

ISSN: 1471-2105