Fast Similarity Search for Protein 3D Structure Databases Using Spatial Topological Patterns

Abstract

It becomes too expensive computationally to compare a query protein with protein structures in a 3D structure databases for determining their similarity. Therefore, we emphasize that solving structural similarity search is to develop fast structure comparison algorithms. We propose a new method for comparing the structural similarity in protein structure databases with a given query protein by using topological pattern of proteins. In our approach, the geometry of SSEs (Secondary Structure Elem ents) is represented by spatial data types and indexed using an Rtree. We discover topological patterns of SSEs in 3D space using 9IM topological relations ac celerated by Rtree index join to all the structures in 3D structure databases. A similarity search algorithm compares topological patterns of a query protein with those of proteins in the structure database. Experimental re sults show that execution time of our method is 3 times faster than DALITE while keeping the accuracy similar. This study identifies that similarity search based on spatial databases can find the similar structures rapidly and generate smal l candidate sets for the generalized alignment tools such as DALI and SSAP.