Protein docking for low-resolution structures.

A typical problem for a docking procedure is how to match two molecules with known 3-D structure so as to predict the configuration of their complex. A very serious obstacle to docking is an inherent inaccuracy in the 3-D structures of the molecules. In general, existing molecular recognition techniques are not designed for cases where (i) conformational changes upon macromolecular complex formation are substantial or (ii) the X-ray data on one or both (macro) molecules are not available, and the structures, based on alternative sources (NMR, modeling), are not well defined. We designed a direct computer experiment using molecules totally deprived of any structural features smaller than 7 A. This was performed on the basis of a previously developed docking algorithm. The modified procedure was applied to a number of known protein complexes taken from the Brookhaven Protein Data Bank. In most cases, a pronounced trend towards the correct structure of the molecular complex was clearly indicated and the real binding sites were predicted. The distinction between the prediction of the antigen-antibody complex and other molecular pairs may reflect important differences in the principles of complex formation. The results strongly suggest the use of our recognition procedure for docking studies where the detailed structures of the molecules are lacking.