Comparing Objects of Different Sizes: Treating Proteins as Strings

For low-resolution force field calculations on proteins, one needs a method to automatically identify weak structural similarities in different molecules. A method is presented which treats sets of three-dimensional coordinates as graphs, and generates a similarity matrix based on local clique matching density that is searched by dynamic programming to identify similarities in the original structures. The method meets the minimal requirements that it works with protein molecules of different sizes and works with gaps and insertions. Examples are given where similarities between protein structures have been detected despite no similarity being detectable by simple inspection.