Algorithms for the protein docking problem usually employ the rigid-body assumption like [Ackermann et al., 1998] or use computationally complex methods of molecular dynamics, see [Lengauer and Rarey, 1996] for an overview. We use a protein representation and docking algorithsm that is based on elastic solids and maximization of surface complemetarity. This allows for changes of the protein surface without time consuming detailled modeling of the atoms and bonds.
A deformation is described by a displacemet vectorfield. Our optmization process finds a deformation that maximizes surface complementarity with minimal internal energy. The detailed description of the is laid out in [Neumann, 1999].
To asses the performance of our approach we tested the system with synthetical test cases representing the three isolated aspects a) translation and b) rotation of the surface patch and c) deformation of isolated areas on the surface. In the rotation test we were able to compensate for rotations of more than 45°.
We are looking for an improved model of conformational changes in the future. We will investigate the possibility of representing the proteins as inhomogeneous and anisotropic bodies. This allows to model hinges where bending is more likely to occur than in areas of decreased flexibility (e.g. dense packing of side chains with very few degrees of freedom). The main part of the computational effort would go into preprocessing of the protein data, without major changes in the optimization process and its computational complexity. Data for the material parameters can be taken from statistical analysis of protein data or using molecular dynamics in the preprocessing step.