FlexOracle: predicting flexible hinges by identification of stable domains

BMC Bioinformatics

Table 1 Summary of predictor results

		GNM	Single-cut predictor (TINKER)	Single-cut predictor (FoldX)	Two-cut predictor
1	Total residues in HAG	13246	13246	13246	13246
2	Test positives	1279	923	292	268
3	Gold Standard positives	152	152	152	152
4	True positives TP (2 ∩ 3)	39	24	14	62
5	False Positives FP (2 – 4)	1240	899	278	206
6	False Negatives FN (3 – 4)	113	128	138	90
7	True Negatives TN (1–2–6)	11854	12195	12816	12888
8	Sensitivity (TP/(TP+FN))	0.26	.157	.092	.41
9	Specificity (TN/(TN+FP))	.9053	.93	.98	.98
10	p-value	8.4·10^-9	1.3·10^-4	6.7·10^-6	3.5·10^-66

"Test positives" for GNM were those residues with first normal mode displacement below 0.02. Recall that the displacements are normalized to range from 0 to 1. For the single-cut predictors, test positives were residues identified as defining a local minimum per the algorithm described in the text. Lastly, "test positives" for the two-cut predictor are those selected under the strict criterion (4-residue window) also described in the text.

ISSN: 1471-2105