ConEVA: a toolbox for comprehensive assessment of protein contacts

BMC Bioinformatics

Table 1 Spearman’s rank correlation coefficient between the length of a protein (L) and evaluation measures for PSICOV predicted long-range contacts in the PSICOV data set. It shows that spread, coverage and X_d are more correlated to L and N_c than precision and mean false positive error, especially below top-L contact selection. For this dataset, the lengths are distributed in the range [50, 266] with mean and standard deviation of 145 and 52 respectively

Contact-Selection	Top-5	Top-L/10	Top-L/5	Top-L/2	Top-L	Top-2 L
L vs Precision	−0.01	−0.07	0.06	0.24	0.26	0.27
L vs Coverage	−0.88	−0.59	−0.51	−0.34	−0.30	−0.31
L vs X_d	0.31	0.35	0.46	0.49	0.51	0.55
L vs FP-Error	−0.05	0.04	−0.06	−0.16	−0.01	0.21
L vs Spread	0.88	0.66	0.60	0.58	0.55	0.57

ISSN: 1471-2105