Statistical analysis of DNA sequences. I

Scaling behavior of nucleotide cluster in DNA sequences

In this paper we study the scaling behavior of nucleotide cluster in 11 chromosomes of Encephalitozoon cuniculi Genome. The statistical distribution of nucleotide clusters for 11 chromosomes is characterized by the scaling behavior of P(S) proportional, variant e(-alphaS), where S represents nucleotide cluster size. The cluster-size distribution P(S(1)+S(2)) with the total size of sequential C-G cluster and A-T cluster S(1)+S(2) were also studied. P(S(1)+S(2)) follows exponential decay. There does not exist the case of large C-G cluster following large A-T cluster or large A-T cluster following large C-G cluster. We also discuss the relatively random walk length function L(n) and the local compositional complexity of nucleotide sequences based on a new model. These investigations may provide some insight into nucleotide cluster of DNA sequence.

Statistical properties of nucleotide clusters in DNA sequences

Using the complete genome of Plasmodium falciparum 3D7 which has 14 chromosomes as an example, we have examined the distribution functions for the amount of C or G and A or T consecutively and non-overlapping blocks of m bases in this system. The function P(S) about the number of the consecutive C-G or A-T content cluster conforms to the relation P(S) proportional, variante(-alphas); values of the scaling exponent alpha(CG) are much larger than alpha(AT); and alpha(AT) of 14 chromosomes are hardly changed, whereas alpha(CG) of 14 chromosomes have a number of fluctuations. We found maximum value of A-T cluster size is much larger than C-G, which implies the existence of large A-T cluster. Our study of the width function xi(m) of cluster C-G content showed that follows good power law xi(m) proportional, variantm(-gamma). The average gamma for 14 chromosomes is 0.931. These investigations provide some insight into the nucleotide clusters of DNA sequences, and help us understand other properties of DNA sequences.

Scaling in nature: from DNA through heartbeats to weather

The purpose of this report is to describe some recent progress in applying scaling concepts to various systems in nature. We review several systems characterized by scaling laws such as DNA sequences, heartbeat rates and weather variations. We discuss the finding that the exponent alpha quantifying the scaling in DNA in smaller for coding than for noncoding sequences. We also discuss the application of fractal scaling analysis to the dynamics of heartbeat regulation, and report the recent finding that the scaling exponent alpha is smaller during sleep periods compared to wake periods. We also discuss the recent findings that suggest a universal scaling exponent characterizing the weather fluctuations.

Scaling features of noncoding DNA

We review evidence supporting the idea that the DNA sequence in genes containing noncoding regions is correlated, and that the correlation is remarkably long range--indeed, base pairs thousands of base pairs distant are correlated. We do not find such a long-range correlation in the coding regions of the gene, and utilize this fact to build a Coding Sequence Finder Algorithm, which uses statistical ideas to locate the coding regions of an unknown DNA sequence. Finally, we describe briefly some recent work adapting to DNA the Zipf approach to analyzing linguistic texts, and the Shannon approach to quantifying the "redundancy" of a linguistic text in terms of a measurable entropy function, and reporting that noncoding regions in eukaryotes display a larger redundancy than coding regions. Specifically, we consider the possibility that this result is solely a consequence of nucleotide concentration differences as first noted by Bonhoeffer and his collaborators. We find that cytosine-guanine (CG) concentration does have a strong "background" effect on redundancy. However, we find that for the purine-pyrimidine binary mapping rule, which is not affected by the difference in CG concentration, the Shannon redundancy for the set of analyzed sequences is larger for noncoding regions compared to coding regions.

Quantification of DNA patchiness using long-range correlation measures

We introduce and develop new techniques to quantify DNA patchiness, and to quantify characteristics of its mosaic structure. These techniques, which involve calculating two functions, alpha(l) and beta(l), measure correlations at length scale l and detect distinct characteristic patch sizes embedded in scale-invariant patch size distributions. Using these new methods, we address a number of issues relating to the mosaic structure of genomic DNA. We find several distinct characteristic patch sizes in certain genomic sequences, and compare, contrast, and quantify the correlation properties of different sequences, including a number of yeast, human, and prokaryotic sequences. We exclude the possibility that the correlation properties and the known mosaic structure of DNA can be explained either by simple Markov processes or by tandem repeats of dinucleotides. We find that the distinct patch sizes in all 16 yeast chromosomes are similar. Furthermore, we test the hypothesis that, for yeast, patchiness is caused by the alternation of coding and noncoding regions, and the hypothesis that in human sequences patchiness is related to repetitive sequences. We find that, by themselves, neither the alternation of coding and noncoding regions, nor repetitive sequences, can fully explain the long-range correlation properties of DNA.

Long-range correlation properties of coding and noncoding DNA sequences: GenBank analysis

An open question in computational molecular biology is whether long-range correlations are present in both coding and noncoding DNA or only in the latter. To answer this question, we consider all 33301 coding and all 29453 noncoding eukaryotic sequences--each of length larger than 512 base pairs (bp)--in the present release of the GenBank to dtermine whether there is any statistically significant distinction in their long-range correlation properties. Standard fast Fourier transform (FFT) analysis indicates that coding sequences have practically no correlations in the range from 10 bp to 100 bp (spectral exponent beta=0.00 +/- 0.04, where the uncertainty is two standard deviations). In contrast, for noncoding sequences, the average value of the spectral exponent beta is positive (0.16 +/- 0.05) which unambiguously shows the presence of long-range correlations. We also separately analyze the 874 coding and the 1157 noncoding sequences that have more than 4096 bp and find a larger region of power-law behavior. We calculate the probability that these two data sets (coding and noncoding) were drawn from the same distribution and we find that it is less than 10(-10). We obtain independent confirmation of these findings using the method of detrended fluctuation analysis (DFA), which is designed to treat sequences with statistical heterogeneity, such as DNA's known mosaic structure ("patchiness") arising from the nonstationarity of nucleotide concentration. The near-perfect agreement between the two independent analysis methods, FFT and DFA, increases the confidence in the reliability of our conclusion.