Skip to main content

Table 1 Sorting of the methods according to the sensitivity and specificity criteria (5 proteins set)

From: Improved prediction of critical residues for protein function based on network and phylogenetic analyses

Method

Proportion of predicted residues

Error

Sen.

Spe.

Union set (10)

38%

0,41

71%

73%

2-connectivity (3)

42%

0.43

74%

67%

Conseq

42%

0,44

76%

67%

Closeness (1)

38%

0.44

68%

71%

Intersection set (9)

38%

0,44

68%

71%

Connectivity (4)

38%

0,45

67%

71%

1/cluster (5)

38%

0.48

63%

70%

Betweenness (2)

38%

0.50

62%

69%

1/eccentricity

45%

0.53

68%

60%

Hssp

38%

0.54

62%

66

  1. For every method, results are obtained in average over five 3D structures: 1HIV, 2LZM, 1BTL, 1A2P and 1GVP. Numbers are obtained the following way: For a given cut-off value (between 0% and 100% of residues that are predicted), we calculate the average (over the five proteins) of the dist(method, cut-off) error, the sensitivity and the specificity. This is done for several cut-off values. Then we choose the cut-off for which the average of the dist(method, cut-off) error is minimal. Note that in most of the cases, the error is the smallest if we predict around 40% of the amino acids to be essential. Furthermore the smallest error for 1/eccentricity is worse than the smallest errors for the five other centrality measurements (1 to 5).