Based　on　the　conclusion　that　different　complexes　have　distinctive　chemo-physical　characters　at　interfaces,　a　specific　scoring　function　was　designed　to　select　the　effective　structures　in　protein-protein　docking　procedure　for　the　Other-type　protein　complexes,　which　are　hard　to　predict.　This　scoring　function　was　composed　of　the　atomic　contact　energy　(E-ACE),　van　der　Waals,　and　electrostatic　interaction　energies.　The　weight　of　each　term　was　obtained　by　the　multiple　linear　regression　approach.　The　test　result　on　17　Other-type　complexes　from　CAPRI　benchmark1　demonstrated　that　the　combinatorial　scoring　function　could　delineate　the　interaction　feature　of　the　Other-type　complexes　and　reflect　the　energy　change　during　the　complex　formation,　and　it　has　certain　capacity　of　discriminating　effective　structures　from　numbers　of　the　docked　modes.　Compared　to　the　residue　pair　potential　(RP),　the　combinatorial　score　could　gain　a　higher　success　rate.　Ranking　the　predicted　models　of　two　targets　in　CARPI　round　8,　the　combinatorial　score　also　exhibits　greater　potential　to　distinguish　the　effective　association　modes.