A New Promoter Recognition Method Based on Features Optimal Selection

Promoter recognition is one of the most significant issues in biology area which plays an important role for analysis of gene regulations. Although currently some research results have been obtained, the accuracy of promoter recognition is still low. In this paper, a new promoter recognition method, named GA-PSO-SVM, which integrates GA, PSO and SVM as a whole is proposed. This method adopts the strategy of selecting promoter features subset and optimizing identification model parameters by turns to select a set of most "informative" or "discriminating" features from the initial features and to get the most optimal identification model for promoter recognition, simultaneously. GA-PSO-SVM is tested on large-scale GenBank DNA sequences. Experiment result shows that our method outperforms several existing best-known methods.