Skip to main content

Table 1 Performance on independent test sets varies with different features

From: Improvement of Dscam homophilic binding affinity throughout Drosophilaevolution

Features

Pearson correlation coefficient (r)

Root mean square error (RMSE)

Initial feature size

Filtered feature size

Composition alone

0.61

0.137

40

35

Composition & exon labels

0.65

0.132

43

25

Pseudo amino acid alone

0.74

0.131

120

105

Pseudo amino acid & exon labels

0.65

0.131

123

80

Composition & pseudo amino acid

0.66

0.119

160

55

Exon labels alone

0.40

0.140

3

3

All three types of features combined

0.75

0.115

163

55

  1. Best correlation coefficient (r) and related root mean squared error (RMSE) for datasets based on the 3 types of features alone and their combinations with 10 fold cross-validations. The feature size indicates the number of features obtained by juxtaposing the different descriptors (initial) and the one after RA algorithm (filtered).