Development of machine learning methods for the discrimination between coding and non-coding conserved sequences