svm PRAT: SVM-based Protein Residue Annotation Toolkit

BMC Bioinformatics

Table 4 Disorder Prediction Performance at CASP8.

Method	ROC	Q_2	S _w
MULTICOM	0.92	0.81	0.61
CBRC-DP_DR	0.91	0.81	0.62
GS-MetaServer2	0.91	0.83	0.66
McGuffin	0.91	0.82	0.64
DISOclust	0.91	0.82	0.64
GeneSilicoMeta	0.90	0.83	0.655
Poodle	0.90	0.80	0.61
CaspIta	0.89	0.78	0.571
fais-server	0.89	0.78	0.56
MULTICOM-CMFR	0.89	0.82	0.64
MARINER*	0.88	0.80	0.61

* - MARINER used svm PRAT to train models for disorder prediction in participation at CASP8 using the kernel with w = f = 11. We used the 723 sequences with disordered residues from the DisPro [7] dataset. The results are the official results from the CASP organizers and were presented by Dr. Joel Sussman at the Weizmann Institute of Science. Q₂ denotes the 2-state accuracy for the prediction and S_wis a weighted accuracy rewarding the prediction of disordered residue.

ISSN: 1471-2105