Abstract
This paper presents feature extraction and estimations of multifractal measures for deoxyribonucleic acid (DNA) sequences, and demonstrates the intriguing possibility of identifying biological functionality using information contained within the DNA sequence. We have developed a technique that seeks patterns or correlations in the DNA sequence at a higher level. The technique has three main steps: (i) transforms the DNA sequence symbols into a modified Lévy walk, (ii) transforms the Lévy walk into a signal spectrum, and (iii) breaks the spectrum into subspectra and treats each of these as an attractor from which the multifractal dimension spectrum is estimated. An optimal minimum window size and volume element size are found for estimation of the multifractal measures. Experimental results show that DNA is a multifractal, and that the multifractality changes depending upon the location (coding or noncoding region) in the sequence.