A COMPUTER VISION-BASED TECHNIQUE FOR 3-D SEQUENCE-INDEPENDENT STRUCTURAL COMPARISON OF PROTEINS

A detailed description of an efficient approach to comparison of prote in structures is presented. Given the 3-D coordinate data of the struc tures to be compared, the system automatically identifies every region of structural similarity between the structures without prior knowled ge of an initial alignment. The method uses the geometric hashing tech nique which was originally developed for model-based object recognitio n problems in the area of computer vision. It exploits a rotationally and translationally invariant representation of rigid objects, resulti ng in a highly efficient, fully automated tool. The method is independ ent of the amino acid sequence and, thus, insensitive to insertions, d eletions and displacements of equivalent substructures between the mol ecules being compared. The method described here is general, identifie s 'real' 3-D substructures and is not constrained by the order imposed by the primary chain of the amino acids. Typical structure comparison problems are examined and the results of the new method are compared with the published results from previous methods. These results, obtai ned without using the sequence order of the chains, confirm published structural analogies that use sequence-dependent techniques. Our resul ts also extend previous analogies by detecting geometrically equivalen t out-of-sequential-order structural elements which cannot be obtained by current techniques.