Using a compressibility measure to distinguish coding and noncoding DNA

Recognition of coding regions is an important phase in gene finding procedures. This paper presents a new method for distinguishing coding and noncoding DNA regions. The proposed method implements compressibility measures that results from Variable Order Markov (VOM) models. In contrast to fixed order Markov models, where the model order is identical for all positions and for all contexts, in VOM models the order may vary-based on a nucleotide position and its contexts. As a result, VOM models are more flexible with respect to model parameterization.