Abstract

Existing machine learning methods for classification of DNA sequence achieve good results, but these methods try to express a DNA sequences as discrete multi-dimensional vector, so when the length of the sequences in the DNA sequence database is not fixed or there exists some omitted characters, these methods can not be used directly. In this paper, we define the new support and growth rate of support to find the frequent emerging patterns from DNA sequence database, and present a classification algorithm FESP based on the frequent emerging sequence patterns. The frequent emerging sequence patterns keep the information provided by the order of bases in gene sequences and can catch interaction among bases. FESP algorithm applies classification rules that are constructed by frequent emerging sequence patterns of each class to classify the new DNA sequences. This method can work on sequences with different lengths or omitted character and shows good performance.