Prediction of neddylation sites from protein sequences and sequence-derived properties

BMC Bioinformatics

Table 2 Effect of different feature sets in selected features on prediction performance.

	5-fold stratified cross-validation^†
Information	Acc	Sp	Sn	MCC	AUC
All selected features	0.91	0.91	0.75	0.44	0.95
without amino acid preferences	0.88	0.89	0.63	0.33	0.88
without amino acid preferences (grouped)	0.89	0.90	0.65	0.35	0.91
without disorder	0.89	0.90	0.72	0.40	0.93
without termini	0.91	0.93	0.68	0.42	0.94
without amino acid occurrence counts	0.91	0.91	0.74	0.44	0.94
without hydrophobicity features	0.91	0.91	0.75	0.44	0.94
without amino acid occurrence ratios	0.91	0.92	0.72	0.43	0.94
without PSSM features	0.91	0.92	0.74	0.44	0.94

^† Cross-validation (CV) results were reported as means of 100 repeats. As two standard errors were not exceeding 0.01, they were not reported.

ISSN: 1471-2105