Identification of novel DNA repair proteins via primary sequence, secondary structure, and homology

BMC Bioinformatics

Table 2 Identification experiments using GO-PDB data

Seq. Similarity	Pos/Neg (% Pos)	Methodolgy	AUC	TPR-0	TPR-1%	TPR-5%
		P-1	0.67	1.32	1.32	23.68
		PS-1–8	0.75	7.89	15.79	36.84
50% (Figure 1)	76/215 (21%)	PF-1	0.53	1.32	5.26	9.21
		BLAST	0.63	0.00	31.58	39.47
		PH-3	0.70	7.89	10.53	30.26
		PSH-3–9	0.64	11.84	18.42	31.58
		P-3	0.68	12.28	14.04	30.70
		PS-1–9	0.77	5.26	7.89	41.23
90%	114/353 (24%)	PF-3	0.64	0.00	7.02	19.30
		BLAST	0.59	0.00	33.33	46.49
		PH-3	0.70	8.77	20.18	39.47
		PSH-3–8	0.74	15.79	19.30	36.84
		P-3	0.96	57.27	76.48	83.12
		PS-3–9	0.96	53.14	64.99	82.76
0%	557/1443 (28%)	PF-3	0.89	25.85	32.50	63.38
		BLAST	0.73	0.00	60.14	66.97
		PH-1	0.90	17.59	25.49	66.07
		PSH-3–8	0.91	39.50	55.30	71.10

The results (including Figure 1) of DNA repair protein identification experiments using proteins drawn from the Protein Data Bank containing the DNA repair Gene Ontology. "Pos/Neg" shows the respective numbers of repair proteins and nuclear non-repair proteins. Column meanings are similar to Figure 1. For each sequence similarity dataset, the top performing method for each metric is underlined (unless a tie exists in more than half of the classifiers).

ISSN: 1471-2105