Multifractal analysis of DNA sequences using a novel chaos-game representation

We present a generalization of the standard chaos-game representation metho d introduced by Jeffrey. To this aim, a DNA symbolic sequence is mapped ont o a singular measure on the attractor of a particular IFS model, which is a perfect statistical representation of the sequence. A multifractal analysi s of the resulting measure is introduced and an interpretation of singulari ties in terms of mutual information and redundancy (statistical dependence) among subsequence symbols within the DNA sequence is provided. The multifr actal spectrum is also shown to be more sensitive for detecting dependence structures within the DNA sequence than the averaged contribution given by redundancy. This method presents several advantages with respect to other r epresentations such as walks or interfaces, which may introduce spurious ef fects. In contrast with the results obtained by other standard methods, her e we note that no general statement can be made on the influence of coding and non-coding content on the correlation length of a given sequence. (C) 2 001 Elsevier Science B.V. All rights reserved.