Prediction of Protein-Protein Binding Interfaces - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 118

ROSETTA-Dock F-measure BEST

0.5

0.49

0.48

0.47

0.46

0.45

ROSETTA-Dock MCC BEST

1

0.98

0.96

0.94

0.92

0.9

Fig. 6.12 Best solutions produced by RosettaDock, ranked according to F-measure ( top ) and
MCC ( bottom ) criteria

top 10 or bottom 10 solutions (usually scoring 0 on either the F-measure or the
MCC scale in the latter case). A more detailed analysis of each complex is required
to determine the reasons behind these extreme values.
The summary presented in Table 6.2 enables us to study the specific features of
each complexation model. Due to the fact that the hydrophobic core model is mark-
edly different from all other software packages, we will focus our analysis on the
best and worst results produced by this model. This decision is also conditioned by
the procedures employed by other toolkits: unlike HADDOCK, RosettaDock or
ZDOCK, the “fuzzy oil drop” model does not generate a large number of candidate
structures, nor does it involve clustering. Instead, it produces a single Δ H pro fi le
which it then uses to pinpoint a specific complexation site. In the case of the “fuzzy
oil drop” model, it is also possible to determine the actual causes of successes and
failures in predicting complexation sites. For these reasons our comparative analysis
will focus on the proteins listed in the first row and the first column of Table 6.2 .