In protein-protein docking prediction system MEGADOCK that we are developing, the target function calculating the docking scores depends on the shape complementarity and electrostatic interaction. And the weight of each term doesn't depend on the target protein and be constant. However, there are a lot of varieties in the shape of proteins and their conformational changes in the docking. Thus, the improvement of the prediction can be expected by changing this proportion in the target function based on the feature of proteins. In this study, we proposed a new method to optimize the electrostatic weight of the target function based on the feature of individual proteins such as the solvent accessible surface area, the surface charge and the structural changes, for improving the docking prediction in the unbound docking.