Classification studies based on a spectral representation of DNA.

The aim of this paper is to provide a new tool to classify the nucleic acid sequences. The profiles based on the values characterizing DNA sequences (descriptors derived from recently introduced graphical representation) are used as a basis of classification schemes related to different aspects of similarity of the sequences. New multicomponent similarity measures based on these descriptors have been defined. Each component of the new measures can be analyzed separately. The new measures are consistent with the standard ones but they contain more detailed information. In particular, it has been shown that within the new approach one can define a quantity which converges to the standard similarity measure. An application of the multicomponent similarity measure to families of histone sequences demonstrates the power and efficiency of the method.