Long-range Correlations in Nucleotide Sequences
C-K. Peng, S.V. Buldyrev, A.L. Goldberger, S. Havlin,
F. Sciortino, M. Simon, and H.E. Stanley
Nature , 356, 168--170 (March 12, 1992)
Abstract
DNA sequences have been analyzed using models, such as an n-step
Markov chain, that incorporate the possibility of short-range
nucleotide correlations. We propose here a method for studying
the stochastic properties of nucleotide sequences by constructing
a 1:1 map of the nucleotide sequence onto a walk, which we term
a "DNA walk". We then use the mapping to provide a quantitative
measure of the correlation between nucleotides over long distances
along the DNA chain. Thus we uncover in the nucleotide sequence
a remarkably long-range power law correlation that implies a new
scale-invariant property of DNA. We find such long-range correlations
in intro-containing genes and in nontranscribed regulatory DNA
sequences, but not in complementary DNA sequences or intron-less genes.
