ABSTRACT Protein–protein docking is an important computational tool for predicting protein–protein interactions. With the rapid development of proteomics projects, more and more experimental binding information ranging from mutagenesis data to three‐dimensional structures of protein complexes are becoming available. Therefore, how to appropriately incorporate the biological information into traditional ab initio docking has been an important issue and challenge in the field of protein‐protein docking. To address these challenges, we have developed a Hybrid DOCKing protocol of template‐based and template‐free approaches, referred to as HDOCK. The basic procedure of HDOCK is to model the structures of individual components based on the template complex by a template‐based method if a template is available; otherwise, the component structures will be modeled based on monomer proteins by regular homology modeling. Then, the complex structure of the component models is predicted by traditional protein‐protein docking. With the HDOCK protocol, we have participated in the CPARI experiment for rounds 28‐35. Out of the 25 CASP‐CAPRI targets for oligomer modeling, our HDOCK protocol predicted correct models for 16 targets, ranking one of the top algorithms in this challenge. Our docking method also made correct predictions on other CAPRI challenges such as protein–peptide binding for 6 out of 8 targets and water predictions for 2 out of 2 targets. The advantage of our hybrid docking approach over pure template‐based docking was further confirmed by a comparative evaluation on 20 CASP‐CAPRI targets. Proteins 2017; 85:497–512. © 2016 Wiley Periodicals, Inc.