Binding　site　prediction　for　protein-protein　complexes　is　a　challenging　problem　in　the　area　of　computational　molecular　biology.　Using　a　set　of　double-chain　complexes　in　Benchmark　3.0,　we　calculated　the　solvent　accessible　surface　areas　and　inter-residue　contact　areas　for　each　monomer　and　propose　a　division　method　of　protein　surface　patches.　We　found　that　the　products　of　the　solvent　accessible　surface　areas　and　internal　contact　areas　of　patches,　the　PSAIA　values,　could　provide　protein　binding　site　information.　In　a　dataset　of　78　complexes,　either　receptors　or　ligands　of　74　complexes　had　interface　patches　with　the　first　or　second　greatest　PSAIA　values　among　all　surface　patches.　A　good　docking　result　was　achieved　when　the　binding　site　information　obtained　with　this　method　was　applied　in　Target　39　of　the　CAPRI　experiment.　This　patch-based　protein　binding　site　prediction　method　differs　from　traditional　methods,　which　are　based　on　single　residue　and　consider　only　surface　residues.　This　provides　a　new　method　for　binding　site　prediction　in　protein-protein　interactions.