Guest lecture on de-novo motif discovery in DNA sequences using parsimonious Markov models

Tapahtuman tyyppi: 
29.11.2011 - 10:15 - 11:00
Ralf Eggeling
Exactum, B119

De-novo motif discovery in DNA sequences using parsimonious Markov models

Computational methods for the prediction of transcription factor binding sites in DNA sequences are necessary for deepening our understanding of gene regulation and thus are of increasing interest in bioinformatics and computational genomics.

We discuss de-novo motif discovery approaches based on promoter models using parsimonious Markov models to take into account statistical dependencies among adjacent nucleotides within the motif. Parsimonious Markov models utilize parsimonious context trees, which generalize well-known context trees from information theory. We discuss expectation-maximization and Gibbs sampling algorithms for this model class, which use efficient dynamic programming algorithms for estimating and sampling parsimonious context trees.

In case studies on ChIP-seq target sequences of the human insulator protein CTCF, we find that the prediction of CTCF binding sites can be improved by using parsimonious Markov models.


Ralf Eggeling is a  PhD student in the group of Prof. Ivo Grosse at the University of Halle, Germany, since he graduated in March 2010 in the field of bioinformatics from the same university. His research interest comprise the biological sequence analysis in general and the identification of functional elements in DNA and RNA in particular.
From computational perspective,  he is nterested in probabilistic graphical models, Bayesian inference, model selection and Bayesian model averaging. He will be visiting Helsinki from Nov. 28 to Dec. 2, hosted by Prof. Petri Myllymäki.

24.11.2011 - 13:07 Petri Myllymäki
24.11.2011 - 13:07 Petri Myllymäki