From Jstacs
Jump to navigationJump to search

The first Jstacs workshop has been held on May, 19th at Martin Luther University. The aim of the workshop is to increase the visibility of Jstacs and the implemented projects, to bring developers and users together, as well as to identify further needs.

Depending on the audience the workshop will be held in English or German.


Time Presenter Type Title & Infos
09:00 - 09:45 Jan Grau & Jens Keilwagen Tutorial Introduction to Jstacs

Information about: Alphabet, AlphabetContainer, Sequence, Sample, Model, ScoringFunction, Classifier, Optimization, Parameter, ParameterSet, StrandModel

09:45 - 10:00 Break
10:00 - 10:45 Jan Grau & Jens Keilwagen Tutorial New Features of Jstacs 1.5

Information about: Alignment, HMM package, Discriminative Sampling

11:00 - 11:30 Lasse Feldhahn Talk Discriminative Metropolis sampler for motif discovery

Abstract: The Metropolis sampler is an instance of MCMC learning, which approximates the distribution of parameters of a model given a learning principle and training data. In this talk, we consider a discriminative Metropolis sampler approximating the supervised posterior. For the evaluation of the objective function, we use existing components of Dispom, a discriminative motif discovery approach implemented in Jstacs. Besides the standard ZOOPS model, these components include a position distribution of motif occurrences. This talk presents the steps necessary for using the model of Dispom for discriminative motif discovery using the Metropolis sampler instead of the previous maximization of the supervised posterior, and presents first results on artificial as well as biological data.

11:30 - 13:00 Lunch
13:00 - 13:30 Ralf Eggeling Talk Maximum Tsallis Entropy Models

Abstract: Classifying short sequence signals is a standard task in genome analysis. One important application is to distinguish splice donor and acceptor sites from decoy. A standard solution of the problem utilizes the maximum entropy principle. Here, we discuss a generalization of this approach based on the Tsallis entropy and its implementation in Jstacs.

13:30 - 14:00 Michael Seifert Talk Exploiting prior knowledge and gene distances in the analysis of tumor expression profiles with extended Hidden Markov Models

Abstract: Changes in gene expression levels play a central role in tumors. Additional information about the distribution of gene expression levels and distances between adjacent genes on chromosomes should be integrated into the analysis of tumor expression profiles. We use a Hidden Markov Model with distance-scaled transition matrices (DSHMM) to incorporate chromosomal distances of adjacent genes on chromosomes into the identification of differentially expressed genes in breast cancer.

14:00 - 14:15 Break
14:15 - 14:45 Michael Scharfe Talk A Hidden Markov Model for Phylogenetic Footprinting

Abstract: Phylogenetic footprinting is an important method to discover overrepresented and cross-species conserved sequence motifs (e.g. TFBSs). An extension of the classical Hidden Markov Model that implements phylogenetic emissions was developed using the HMM-Package of the current Jstacs version.

14:45 - 15:15 Martin Gleditzsch Talk PhyloBayes: An approach to model simultaneously evolutionary dependencies and statistical dependencies between neighbored DNA-Positions with Bayesian Networks

Abstact: Two approaches exist in the field of predicting regulatory elements in a set of sequences. The first approach called de-novo motif discovery uses information about dependencies within binding sites but neglects information about evolutionary conservation. The second approach called phylogenetic footprinting uses information about evolutionary conservation but neglects information about dependencies within binding sites. Here, we present the fusion of these two complementary approaches using Bayesian networks and its integration in Jstacs.

15:15- 15:30 Break
15:30 - 16:30 Discussion Discussion on future directions and improvements of Jstacs


The workshop has been held on May, 19th in room 5.09 at the Institute of Computer Science at Martin Luther University Halle (Von-Seckendorff-Platz 1).


If you have any questions or comments, please do not hesitate to contact us by mail