Mutual Information for Examining Correlations in Dna

Abstract

This paper examines two methods for finding whether long-range correlations exist in DNA: a fractal measure and a mutual information technique. We evaluate the performance and implications of these methods in detail. In particular we explore their use comparing DNA sequences from a variety of sources. Using software for performing in silico mutations, we also consider evolutionary events leading to long range correlations and analyse these correlations using the techniques presented. Comparisons are made between these virtual sequences, randomly generated sequences, and real sequences. We also explore correlations in chromosomes from different species.

Information theory reveals large-scale synchronisation of statistical correlations in eukaryote genomes

“…The correlation structure of a DNA sequence has been extensively studied within the last decade, particularly with methods from information theory (see, e.g., Herzel et al, 1994;Holste et al, 2000;Grosse et al, 2002;Berryman et al, 2004). In these investigations, the mutual information function has turned out to reveal important features of the information content of DNA sequences.…”

“…In many cases the biological origin of such correlations is still far from being fully understood (see, e.g., Ouyang et al, 2004;Garte, 2004 for recent studies). The bulk of scientific efforts to apply information theory as well as various other statistical methods to studying DNA sequences, however, focusedand still focuses (Berryman et al, 2004)-on long-range correlations (Voss, 1992;Li and Kaneko, 1992;Peng et al, 1992;Karlin and Brendel, 1993;Buldyrev et al, 1995). Large-scale statistical properties of DNA sequences have recently received a lot of attention in the statistical physics community.…”

Data Compression Concepts and Algorithms and Their Applications to Bioinformatics

“…Bauer and colleagues show that the average mutual information profile can be used to cluster subtypes of the HIV-1 virus. The fact that average mutual information profiles reflect evolutionary relationships has been demonstrated by Berryman et al [31] who show how the average mutual information profile for a chromosome reflects events in the organisms evolutionary history. Holste et al [32] demonstrate the effect of evolutionary history on the average mutual information profile by showing the effect of evolutionary events on particular characteristics of the profile for various human chromosomes.…”

Use of average mutual information for studying changes in HIV populations

“…Developed by Shannon [1] for the analysis of communication systems, it has been used in a variety of applications in the biological fields. It has been used to examine covariation of different sites in the V3 loop of the HIV genome [2], to investigate correlations between sites in protein sequences[3], [4], [5], and to differentiate between coding and noncoding regions[6], to investigate long range correlations [7], to develop species signatures [8], for fragment assembly [9] to study coevolving sites in polypeptide sequences [2], [10], for secondary structure prediction [11], [12], and to study relationships between genes and their phenotypes [13]. …”