soluble). Meanwhile, docking of two soluble proteins has to be
careful and well validated.
Despite these difficulties, protein–protein docking is currently a
well-established method, albeit not as widely used as small-
molecular docking [8]. The number of reports on protein–protein
docking applications are growing at an extraordinary rate. It can be
explained by the emerging enormous importance of PPIs in many
fields. The huge number of reports makes a detailed review impos-
sible to fit in this chapter. However, some of the most recent reports
may serve as good examples of various strategies of input data
preparation, docking, and analysis of the results.2 Basic Principles of Protein–Protein Docking
Although it could be expected that the performance of protein–-
protein docking is far below the one of small-molecular docking
[8], an excellent review by Joe ̈l Janin [11] clearly does not support
this assumption. Moreover, The Critical Assessment of PRedicted
Interactions (CAPRI) [10–13], a community-wide initiative estab-
lished in 2001, regularly reports the considerable progress in pro-
tein–protein docking algorithms and more and more reliable
protein complexes structure prediction [14]. Among 42 CAPRI
protein–protein complexes modeled in 2001–2010 only six have
not been successfully modeled.
Docking of two proteins requires appropriate rotational and
translational positioning of the binding partners, which can be
efficiently solved by fast-Fourier transformation techniques or geo-
metric hashing techniques, both referred to as rigid-body docking
[15]. Acommon assumption of protein–protein docking methods is
that protein–protein interaction can be modeled based on shape
complementarity and using simplified amino acid models (in a simi-
lar fashion as typically used in early day molecular mechanics united-
atom force fields) [8]. Rigid-body docking programs usually locate
one protein in a fixed position while the other is moved, exploringits
rotatranslational space around the first one [16]. The process of
docking usually consists of two or three steps [17] which include:
(a) the searching stage which involves generation of putative pro-
tein–protein complexes, (b) the sampling stage which involves clus-
tering andscoring of complexesobtained in stage (a), and (c)the last
stage which includes refinement (optional) and ranking of potential
solutions (scoring, filtering) [16]. In the scoring step, the candidate
protein complexes are evaluated using scoring functions reflecting
the geometric and physicochemical complementarity of the pro-
teins, using measures that describe electrostatics, H-bonding inter-
actions and solvation, and that might include also empirical
potentials or database-derived functions [16].
While originally most of the protein–protein docking methods
were based on the rigid-docking approach, the majority of theProtein-Protein Docking 287