FROG (Fitted Rotation and Orientation of protein structure by means of real-coded Genetic algorithm): Asynchronous Parallelizing for Protein Structure-Based Comparison on the Basis of Geometrical Similarity

摘要

A generic protein structure-based comparison method is able to fully retrieve biological interest from large-scaled ample data sets. Unlike other methods hitherto proposed, a generic method needs neither initial equivalent pairs nor scoring matrices.Although our purpose is to propose such a generic method, structure alignment patterns are firstly considered in this study; global alignment uses linearly ordered residue-residue equivalent pairs along chain directions. Linear local alignment also usesordered fragment-fragment equivalences following chain directions, while non-linear local alignment adopts disordered fragment-fragment matches. The other is to find linear and/or non-linear fragments without keeping chain directions. All possible occurrences of the patterns have unique aspects in the comparisons, thus these have to be emphasized separately and simultaneously. In this study, we design a real-coded Genetic Algorithm, GA, improving deficits of Simple GA in the filed of protein structure-based comparison. Moreover, Ninf asynchronously parallelizes the GA. The proposed GA clearly infers conforrnational relationships between homologous and analogous protein families using simultaneous optimization of local similarity and global superposition.