Skip to main content
. 2008 Jul 1;24(13):i190–i195. doi: 10.1093/bioinformatics/btn166

Table A1.

Performances obtained with random splitting of the cross-validation sets

Method Accuracy (%) Correlation Sensitivity (%)
Specificity (%)
+ +
L20 93 0.86 93 93 93 93
L400 97 0.94 97 97 97 97

L20 is the SVM trained with the residue composition. L400 is trained with the difference in dipeptide composition of the sequences. Symbols + and − indicate increased and decreased thermostability classes, respectively.