HADDOCK is a protein-protein docking method that uses biochemical and biophysical interaction data, such as NMR titration or mutagenesis data, to guide the docking process. These data are introduced as ambiguous interaction restraints (AIRs) to define the intermolecular distances between residues involved in the interaction. The accuracy of HADDOCK is demonstrated with three molecular complexes: the EIN-HPr complex, the E2A-HPr complex, and the gp120-CD4 complex. In all cases, the best structures generated by HADDOCK, which have the lowest intermolecular energy, are closest to the published structures of the complexes. The method uses three stages: (i) randomization of orientations and rigid body energy minimization, (ii) semirigid simulated annealing in torsion angle space, and (iii) final refinement in Cartesian space with explicit solvent. During these stages, the side chains at the interface are allowed to move to optimize the interface packing. The results show that HADDOCK can generate accurate docking solutions starting from either the complex or free form structures. The method is robust and can be used with various types of experimental data, including mutagenesis data, to define ambiguous interaction restraints. The use of AIRs allows HADDOCK to search through all possible configurations around the interacting site and find the most favorable pair of interacting amino acids. The method has been validated using NMR and crystallographic data, and the results show a good correlation between the intermolecular energy and the iRMSD from the target structure. The HADDOCK approach is a powerful tool for protein-protein docking and can be used to improve the accuracy of structural predictions.HADDOCK is a protein-protein docking method that uses biochemical and biophysical interaction data, such as NMR titration or mutagenesis data, to guide the docking process. These data are introduced as ambiguous interaction restraints (AIRs) to define the intermolecular distances between residues involved in the interaction. The accuracy of HADDOCK is demonstrated with three molecular complexes: the EIN-HPr complex, the E2A-HPr complex, and the gp120-CD4 complex. In all cases, the best structures generated by HADDOCK, which have the lowest intermolecular energy, are closest to the published structures of the complexes. The method uses three stages: (i) randomization of orientations and rigid body energy minimization, (ii) semirigid simulated annealing in torsion angle space, and (iii) final refinement in Cartesian space with explicit solvent. During these stages, the side chains at the interface are allowed to move to optimize the interface packing. The results show that HADDOCK can generate accurate docking solutions starting from either the complex or free form structures. The method is robust and can be used with various types of experimental data, including mutagenesis data, to define ambiguous interaction restraints. The use of AIRs allows HADDOCK to search through all possible configurations around the interacting site and find the most favorable pair of interacting amino acids. The method has been validated using NMR and crystallographic data, and the results show a good correlation between the intermolecular energy and the iRMSD from the target structure. The HADDOCK approach is a powerful tool for protein-protein docking and can be used to improve the accuracy of structural predictions.