Designing of a novel GA based on fuzzy system for prediction of CpG islands in the human genome

In this paper we proposed a novel genetic algorithm based on fuzzy system for identification CpG islands in human genome, called FGA-CGI (fuzzy GA-CpG Island). CpG islands play a fundamental role in genome analysis and annotation and contribute to increase the accuracy of promoter prediction. Recently, some approaches rely on large parameter space algorithms of predicting the CpG islands have been proposed in the literature. The goal of our proposed method was that using the evolutionary algorithms with fuzzy system and machine learning to identify CpG islands. A fuzzy expert system was implemented to dynamically adapt the crossover rate and mutation rate in GA for identify significant of CpG islands in human genome, and reinforcement learning serve as extend operation for combined the best subset of islands. In this study, three public tools for identification CpG islands were used to compare with FGA-CGI for the assessment of five prediction performance and statistically analysis. Experimental results reveal that our method can adjust the two variables to escape local optimal by fuzzy system and identify more number of CpG islands. In addition, FGA-CGI had capable of higher performance and precisely predicting statistically significant CpG islands in target sequences than these previous tools.