Skip to main content
Fig. 5 | BioData Mining

Fig. 5

From: Application of an interpretable classification model on Early Folding Residues during protein folding

Fig. 5

The Classification Correlation Matrix for the classification of Early Folding Residues. a The CCM depicts the positive impact of individual features for the classification performance on its main diagonal. Especially, PlipHpCL, LF, and PlipBN are features which discriminate EFR and LFR. The influence of ordered secondary structure elements was shown before [17, 19]. Both betweenness and closeness centrality tend to be increased for EFR which indicates their importance for the assembly of secondary structure elements by long-range hydrophobic interactions [17]. Other entries of the matrix describe pairs of features which are positively (red) or negatively (blue) correlated and increase classification performance further. b The standard correlation matrix of all features of the whole dataset. Again, positive and negative correlations are depicted in red and blue respectively. Interestingly, the features pointed out by GMLVQ do not stand out. Vice versa, strong correlations between features do not imply a favorable influence on the classification performance

Back to article page