Berlin 2005 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

AKB: Biologische Physik

AKB 100: Poster Session I

AKB 100.62: Poster

Samstag, 5. März 2005, 16:45–18:45, Poster TU D

Information theory reveals large-scale synchronisation of statistical correlations in Eukaryote genomes — •Manuel Dehnert¹, Werner E. Helm², and Marc-Thorsten Hütt¹ — ¹Bioinformatics Group, Department of Biology, Darmstadt University, Schnittspahnstr. 3-5, D-64287 Darmstadt — ²Mathematics and Science Faculty, University of Applied Sciences, D-64295 Darmstadt

We study short-range correlations in DNA sequences with methods from information theory and statistics. We find a persisting degree of identity between the correlation patterns of different chromosomes of a species. Except for the case of human and chimpanzee inter-species differences in this correlation pattern allow robust species distinction: In a clustering tree based upon the correlation curves on the level of individual chromosomes distinct clusters for the individual species are found. This capacity of distinguishing species persists, even when the length of the underlying sequences is drastically reduced. In comparison to the standard tool for studying symbol correlations in DNA sequences, namely the mutual information function, we find that an autoregressive model for higher-order Markov processes significantly improves species distinction due to an implicit subtraction of random background.