A new computationally efficient and automated "soft docking" algorithm is d
escribed to assist the prediction of the mode of binding between two protei
ns, using the three-dimensional structures of the unbound molecules. The me
thod is implemented in a software package called BiGGER (Bimolecular Comple
x Generation with Global Evaluation and Ranking) and works in two sequentia
l steps: first, the complete 6-dimensional binding spaces of both molecules
is systematically searched. A population of candidate protein-protein dock
ed geometries is thus generated and selected on the basis of the geometric
complementarity and amino acid pairwise affinities between the two molecula
r surfaces. Most of the conformational changes observed during protein asso
ciation are treated in an implicit way and test results are equally satisfa
ctory, regardless of starting from the bound or the unbound forms of known
structures of the interacting proteins. In contrast to other methods, the e
ntire molecular surfaces are searched during the simulation, using absolute
ly no additional information regarding the binding sites. In a second step,
an interaction scoring function is used to rank the putative docked struct
ures. The function incorporates interaction terms that are thought to be re
levant to the stabilization of protein complexes. These include: geometric
complementarity of the surfaces, explicit electrostatic interactions, desol
vation energy, and pairwise propensities of the amino acid side chains to c
ontact across the molecular interface. The relative functional contribution
of each of these interaction terms to the global scoring function has been
empirically adjusted through a neural network optimizer using a learning s
et of 25 protein-protein complexes of known crystallographic structures. In
22 out of 25 protein-protein complexes tested, near-native docked geometri
es were found with C-alpha RMS deviations less than or equal to 4.0 Angstro
m from the experimental structures, of which 14 were found within the 20 to
p ranking solutions. The program works on widely available personal compute
rs and takes 2 to 8 hours of CPU time to run any of the docking tests herei
n presented. Finally, the value and limitations of the method for the study
of macromolecular interactions, not yet revealed by experimental technique
s, are discussed. (C) 2000 Wiley-Liss, Inc.