Structural superposition of proteins with unknown alignment and detection of topological similarity using a six-dimensional search algorithm

An algorithm for the rigid-body superposition of proteins is described and tested. No prior knowledge of equivalent residues is required. To find the common structural core of two proteins, an exhaustive grid search is conducted in three-dimensional angle space, and at each grid point a fast translation search in three-dimensional space is performed. The best superposition at a given angle set is defined by that translation vector which maximizes the weighted number of equivalent C alpha atoms. Filters using the information about the sequential character of the polypeptide chain are employed to identify that rotation and translation which yields the highest topological similarity of the two proteins. The algorithm is shown to find the best superposition of distantly related structures, and to be capable of finding similar structures to a given atomic model in the Brookhaven Protein Data Bank. In a search using granulocyte-macrophage colony-stimulating factor as a template, all other four-helix bundle cytokines with up-up-down-down topology were found to give the highest values of a topological similarity score, followed by interferon-beta and -gamma and those four-helix bundles with the more common up-down-up-down topology. In another example, the insertion domain of the long variant adenylate kinases is demonstrated to share its fold with rubredoxin.