Figure 2.

The SVM classifier performance on single features and on various feature combinations. N-gram features (n = 1..5) provide compositional information, secondary structure (ss2), solvent accessibility (acc), disorder (vsl2) and low-complexity features (seg) provide structural information, and positional specific similarity matrix (pssm) that provides evolutionary information, are used in different combination to identify a set of features that can provide best classification performance.

Wang et al. BMC Bioinformatics 2011 12:251   doi:10.1186/1471-2105-12-251
Download authors' original image