svm PRAT: SVM-based Protein Residue Annotation Toolkit

BMC Bioinformatics

Table 2 Classification Performance on the Disorder Dataset.

w	f= 1		f= 3		f= 5		f= 7		f= 9		f= 11
	ROC	F1	ROC	F1	ROC	F1	ROC	F1	ROC	F1	ROC	F1
3	0.775	0.312	0.800	0.350	-	-	-	-	-	-	-	-
7	0.815	0.366	0.817	0.380	0.816	0.384	0.816	0.383	-	-	-	-
11	0.821	0.378	0.826	0.391	0.828	0.396	0.826	0.400	0.824	0.404	0.823	0.403
13	0.823	0.384	0.829	0.398	0.832*	0.405	0.830	0.404	0.828	0.407	0.826	0.409
3	0.811	0.370	0.811	0.369	-	-	-	-	-	-	-	-
7	0.845	0.442	0.849	0.450	0.848	0.445	0.845	0.442	-	-	-	-
11	0.848	0.464	0.855	0.478	0.858	0.482	0.858	0.480	0.855	0.470	0.853	0.468
13	0.848	0.473	0.855	0.484	0.859	0.490	0.861*	0.492	0.860	0.487	0.857	0.478
3	0.815	0.377	0.816	0.379	-	-	-	-	-	-	-	-
7	0.847	0.446	0.852	0.461	0.852	0.454	0.851	0.454	-	-	-	-
11	0.848	0.469	0.856	0.482	0.860	0.491	0.862	0.491	0.861	0.485	0.862	0.485
13	0.847	0.473	0.856	0.485	0.861	0.491	0.864	0.495	0.865*	0.494	0.864	0.492
3	0.836	0.418	0.838	0.423	-	-	-	-	-	-	-	-
7	0.860	0.472	0.862	0.476	0.860	0.473	0.859	0.468	-	-	-	-
11	0.861	0.490	0.867	0.496	0.868	0.498	0.868	0.495	0.866	0.488	0.865	0.485
13	0.860	0.497	0.867	0.503	0.870	0.503	0.871*	0.503	0.870	0.498	0.868	0.492
3	0.842	0.428	0.841	0.428	-	-	-	-	-	-	-	-
7	0.869	0.497	0.870	0.499	0.869	0.494	0.867	0.489	-	-	-	-
11	0.871	0.516	0.875	0.518	0.877	0.517	0.877	0.512	0.874	0.508	0.873	0.507
13	0.869	0.519	0.875	0.522	0.878	0.521	0.879**	0.519	0.879	0.518	0.876	0.514

DISPro [7] reports a ROC score of 0.878. The numbers in bold show the best models for a fixed w parameter, as measured by ROC. , ℬ, and represent the PSI-BLAST profile, BLOSUM62, and YASSPP scoring matrices, respectively. soe, rbf, and lin represent the three different kernels studied using the W_{w, f}as the base kernel. * denotes the best classification results in the sub-tables, and ** denotes the best classification results achieved on this dataset. For the best model we report a Q₂ accuracy of 84.60% with an se rate of 0.33.

ISSN: 1471-2105