A genetic algorithm taking account of substructures for NMR three-dimensional protein structure determination

Nuclear magnetic resonance (NMR) spectroscopy is a promising technique for the three-dimensional structure determination of proteins that is one of the most important problems in post-sequence era. This technique has a serious problem that it takes several months for an expert to analyze the data of only one protein. In order to remedy the problem, Ono et al. (2002) have proposed an automatic method based on a genetic algorithm (GA) for analyzing the data and determining the three-dimensional structures of proteins and reported that they had good results on relatively small-scale problems. In this paper, to get good results on larger-scale problems, we propose a new initial population generation method and the substructure exchange crossover (SSXX) that inherits substructures as characters from parents to offspring. In order to examine the effectiveness of the proposed method, we perform some experiments.