Fractal landscape analysis of DNA walks

By mapping nucleotide sequences onto a "DNA walk", we uncovered remarkably long-range power law correlations [Nature 356 (1992) 168] that imply a new scale invariant property of DNA. We found such long-range correlations in intron-containing genes and in non-transcribed regulatory DNA sequences, but not in cDNA sequences or intron-less genes. In this paper, we present more explicit evidences to support our findings.     

Statistical mechanics in biology: how ubiquitous are long-range correlations?

The purpose of this opening talk is to describe examples of recent progress in applying statistical mechanics to biological systems. We first briefly review several biological systems, and then focus on the fractal features characterized by the long-range correlations found recently in DNA sequences containing non-coding material. We discuss the evidence supporting the finding that for sequences containing only coding regions, there are no long-range correlations. We also discuss the recent finding that the exponent alpha characterizing the long-range correlations increases with evolution, and we discuss two related models, the insertion model and the insertion-deletion model, that may account for the presence of long-range correlations. Finally, we summarize the analysis of long-term data on human heartbeats (up to 10(4) heart beats) that supports the possibility that the successive increments in the cardiac beat-to-beat intervals of healthy subjects display scale-invariant, long-range "anti-correlations" (a tendency to beat faster is balanced by a tendency to beat slower later on). In contrast, for a group of subjects with severe heart disease, long-range correlations vanish. This finding suggests that the classical theory of homeostasis, according to which stable physiological processes seek to maintain "constancy," should be extended to account for this type of dynamical, far from equilibrium, behavior.     

Long-range power-law correlations in condensed matter physics and biophysics

We discuss the appearance of long-range power-law correlations in various systems of interest to condensed matter physicists and biophysicists, with emphasis on the recent discovery of long-range correlations in DNA sequences that contain non-coding regions.     

Long range correlations in DNA: scaling properties and charge transfer efficiency

We address the relation between long-range correlations and charge transfer efficiency in aperiodic artificial or genomic DNA sequences. Coherent charge transfer through the highest occupied molecular orbital states of the guanine nucleotide is studied using the transmission approach, and the focus is on how the sequence-dependent backscattering profile can be inferred from correlations between base pairs.     

Statistical and linguistic features of noncoding DNA: A heterogeneous «Complex system»

Summary We present evidence supporting the idea that the DNA sequence in genes containingnoncoding regions is correlated, and that the correlation is remarkably long range-indeed, base pairsthousands of base pairs distant are correlated. We do not find such a long-range correlation in the coding regions of the gene; we utilize this fact to build aCoding Sequence Finder algorithm, which uses statistical ideas to locate the coding regions of an unknown DNA sequence. We resolve the problem of the ?non-stationarity? feature of the sequence of base pairs (that the relative concentration of purines and pyrimidines changes in different regions of the mosaic-like chain) by describing a new algorithm calledDetrended Fluctuation Analysis (DFA). We address the claim of Voss that there is no difference in the statistical properties of coding and noncoding regions of DNA by systematically applying the DFA algorithm, as well as standard FFT analysis, to every DNA sequence (33 301 coding and 29 453 non-coding) in the entire GenBank database. We describe a simple model to account for the presence of long-range power law correlations (and the systematic variation of the scaling exponent α with evolution) which is based upon a generalization of the classic Lévy walk. Finally, we describe briefly some recent work showing that thenoncoding sequences have certain statistical features in common with natural languages. Specifically, we adapt 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. We suggest that noncoding regions in eukaryotes may display a smaller entropy and larger redundancy than coding regions for plants and invertebrates, further supporting the possibility that noncoding regions of DNA may carry biological information. Paper presented at the I International Conference on Scaling Concepts and Complex Fluids, Copanello, Italy, July 4–8, 1994.     

Solvable sequence evolution models and genomic correlations

We study a minimal model for genome evolution whose elementary processes are single site mutation, duplication and deletion of sequence regions, and insertion of random segments. These processes are found to generate long-range correlations in the composition of letters as long as the sequence length is growing; i.e., the combined rates of duplications and insertions are higher than the deletion rate. For constant sequence length, on the other hand, all initial correlations decay exponentially. These results are obtained analytically and by simulations. They are compared with the long-range correlations observed in genomic DNA, and the implications for genome evolution are discussed.     

Current-voltage characteristics of double-strand DNA sequences

We use a tight-binding formulation to investigate the transmissivity and the current-voltage (I-V) characteristics of sequences of double-strand DNA molecules. In order to reveal the relevance of the underlying correlations in the nucleotides distribution, we compare the results for the genomic DNA sequence with those of artificial sequences (the long-range correlated Fibonacci and Rudin-Shapiro one) and a random sequence, which is a kind of prototype of a short-range correlated system. The random sequence is presented here with the same first neighbors pair correlations of the human DNA sequence. We found that the long-range character of the correlations is important to the transmissivity spectra, although the I-V curves seem to be mostly influenced by the short-range correlations.     

Statistical Dynamics of Clustering in the Genome Structure

Clustering and long-range correlations in the nucleotide sequences of different categories of organisms are studied. As a result of clustering, the size distribution of coding and non-coding DNA regions is estimated analytically using the Generalised Central Limit Theorem.The alternation of coding regions (which follow a short range size distribution) with non-coding regions (which follow a long range size distribution in higher organisms) leads to DNA structures which have a striking resemblance to random Cantor Fractals. For lower organisms (such as viruses, procaryotes etc.) long-range correlations are sporadically observed and the DNA structures do not present fractality.Statistical models are proposed based on biologically motivated dynamical mechanisms (such as aggregation of oligonucleotides, influx and DNA length reduction), which can account for the above statistical features.     

Nucleotide composition effects on the long-range correlations in human genes

We use the wavelet transform to investigate the fractal scaling properties of coding and noncoding human DNA sequences. We find that the strength of the long-range correlations observed in the introns increases with the guanine-cytosine (GC) content, while coding sequences show no such correlations at any GC content. However, we demonstrate that long-range correlations can be detected when the coding sequences are undersampled by retaining the third base of each codon only. This strongly suggests that the observed correlations are not likely to be due to insertion-deletion mechanisms. We comment about the origin of these correlations in terms of putative dynamical processes that could produce the isochore structure of the human genome. Received: 18 August 1997 / Accepted: 29 October 1997     

Statistical methods in nonlinear dynamics

Sensitivity to initial conditions in nonlinear dynamical systems leads to exponential divergence of trajectories that are initially arbitrarily close, and hence to unpredictability. Statistical methods have been found to be helpful in extracting useful information about such systems. In this paper, we review briefly some statistical methods employed in the study of deterministic and stochastic dynamical systems. These include power spectral analysis and aliasing, extreme value statistics and order statistics, recurrence time statistics, the characterization of intermittency in the Sinai disorder problem, random walk analysis of diffusion in the chaotic pendulum, and long-range correlations in stochastic sequences of symbols.     

New class of level statistics in correlated disordered chains

We study the properties of the level statistics of 1D disordered systems with long-range spatial correlations. We find a threshold value in the degree of correlations below which in the limit of large system size the level statistics follows a Poisson distribution (as expected for 1D uncorrelated-disordered systems), and above which the level statistics is described by a new class of distribution functions. At the threshold, we find that with increasing system size, the standard deviation of the function describing the level statistics converges to the standard deviation of the Poissonian distribution as a power law. Above the threshold we find that the level statistics is characterized by different functional forms for different degrees of correlations.     

Statistical Dynamics of Clustering in the Genome Structure

Clustering and long-range correlations in the nucleotide sequences of different categories of organisms are discussed. Clustering, mostly observed in higher eucaryotes, can be found at different length scales in DNA and Central Limit Theorems are used as links between these length scales. Several dynamical, statistical, mean-field models are proposed based on biologically motivated dynamical mechanisms and they successfully reproduce both the short range behavior observed in coding DNA and the long range, out-of-equilibrium features of non-coding DNA. Such dynamical mechanisms include aggregation of oligonucleotides, influx and DNA length reduction schemes, transpositions, and fusions of large DNA macromolecules. Fractality can be inferred from the short and long range correlations observed in the sequence structure of higher eucaryotes, where the non-coding part is relatively extended. In these organisms the DNA coding/non-coding alternation has the characteristics of finite size, fractal, random sets.     

Limb dominance changes in walking evolution explored by asymmetric correlations in gait dynamics

Fluctuations in the stride interval time series of unconstrained walking are not random but seem to exhibit long-range correlations that decay as a power law (Hausdorff et al. (1995) [35]). Here, we examine whether asymmetries are present in the long-range correlations of different gait parameters (stride, swing and stance intervals) for the left and right limbs. Gait dynamics corresponding to 16 healthy subjects were obtained from the Physionet database, which contains stride, stance and swing intervals for both left and right limbs. Detrended Fluctuation Analysis (DFA) revealed the presence of asymmetric long-range correlations in all gait cycle variables investigated. A rich variety of scaling exponent dynamics was found, with the presence of synchronicity, decreased correlations and dominant correlations. The results are discussed in terms of the hypothesis that reduced strength of long-range correlations reflect both enhanced stability and adaptability.     

DNA序列的核苷酸统计关联

探索核苷酸统汁关联的规律性是遗传语言研究的基础。本文评述了这个领域的工作,着重讨论了核苷关联的短程为主性,DNA序列的信息参数分析,以及关联的进化相关性。文中还强调了核苷关联的生物学意义及这一研究的可能生物学应用,其中包括:构建进化树,预测蛋白质二级结构,寻求碱基关联偏好模的规律性,导出编码序列和非编码序列的遗传语言差别,通过关联谱和偏好模发现阅读框架,研究核苷关联和基因表达的关系(表达增强网络),研究长周期关联及功率谱的低频行为等。     

Fractals in DNA sequence analysis

Fractal methods have been successfully used to study many problems in physics,mathematics,engineering,finance,and even in biology,There has been an increasing interest in unravelling the mysteries of DNA;for example,how can we distinguish coding and noncoding sequences,and the problems of classification and evolution relationship of organisms are key problems in bioinformatics,Although much research has been carried out by taking into consideration the long-range correlations in DNA sequences,and the global fractal dimension has been used in these works by other people,the models and methods are somewhat rough and the results are not satisfactory.In recent years,our group has introduced a time series model(statistical point of view)and a visual representation (geometrical point of view) to DNA sequence analysis.We have also used fractal dimension,correlation dimension,the Hurst exponent and the dimension spectrum (multifractal analysis)to discuss problems in this field.In this paper,we introduce these fractal models and methods and the results of DNA sequence analysis.     

电子商务中人类活动的标度行为实证研究

应用去趋势波动分析法,对电子商务中人类网上购物行为进行研究,首次探讨了人类浏览及购买行为时间序列(数量波动)标度律.首先,研究发现人类网上购物行为呈现出明显的周期性,其时间序列的概率密度函数具有显著的双模态特征.其次,利用傅里叶变换方法分析浏览以及购买行为时间序列的功率谱,发现其演化过程不同于无关联的泊松过程.最后,基于功率谱过滤周期性趋势的影响,对去除周期趋势后的浏览和购买行为时间序列进行去趋势波动分析,发现其标度行为表明其具有较强的长程关联特性,且平均标度值近似为1,表明其具有自组织临界性.实证研究结果与其他领域如因特网交通流和金融市场价格波动的标度行为相似,有助于理解人类活动如何影响电子商务系统演化和提高在线商务活动效率,对分析电子商务中人类行为活动的机制和预测其波动趋势具有重要的启示作用.     

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.     

Hamiltonian and Brownian systems with long-range interactions: III. The BBGKY hierarchy for spatially inhomogeneous systems

We study the growth of correlations in systems with weak long-range interactions. Starting from the BBGKY hierarchy, we determine the evolution of the two-body correlation function by using an expansion of the solutions of the hierarchy in powers of 1/N in a proper thermodynamic limit N→+∞, where N is the number of particles. These correlations are responsible for the “collisional” evolution of the system beyond the Vlasov regime due to finite N effects. We obtain a general kinetic equation that can be applied to spatially inhomogeneous systems and that takes into account memory effects. These peculiarities are specific to systems with unshielded long-range interactions. For spatially homogeneous systems with short memory time like plasmas, we recover the classical Landau (or Lenard-Balescu) equations. An interest of our approach is to develop a formalism that remains in physical space (instead of Fourier space) and that can deal with spatially inhomogeneous systems. This enlightens the basic physics and provides novel kinetic equations with a clear physical interpretation. However, unless we restrict ourselves to spatially homogeneous systems, closed kinetic equations can be obtained only if we ignore some collective effects between particles. General exact coupled equations taking into account collective effects are also given. We use this kinetic theory to discuss the processes of violent collisionless relaxation and slow collisional relaxation in systems with weak long-range interactions. In particular, we investigate the dependence of the relaxation time with the system size N and try to provide a coherent discussion of all the numerical results obtained for these systems.     

Nonextensive entropy and the non-Gaussian distribution of magnetization in two-dimensional spin ice

The statistical properties of the magnetization of the finite clusters of two-dimensional spin ice have been investigated. It has been shown by Monte Carlo simulations that the short-range ice rules in two dimensions lead to long-range correlations, which decay by a power law with distance. The long-range correlations, in turn, cause the nonextensivity of entropy and inapplicability of the central limit theorem for the magnetization. The behavior of the moments and distribution function of the magnetization with the cluster size disagrees with the theoretical predictions of the dipolar behavior of the correlation functions in two-dimensional spin ice.