News, current and recent activities

Research

Regulatory Networks

The rapid advances in measurement technologies pose both a chance and a challenge to biology. In order to obtain a systems level understanding of the cell, we develop and apply new statistical models that are capable of dealing with vast amounts of heterogeneous data. These models have to be robust against measurement error, they have to cope with the n<p-situation (relatively few samples compared to the number of variables), and, most importantly, they must lead to interpretable predictions.
The class of probabilistic graphical models provides a very general framework which is well suited to this situation. Our goal is to specialise these models in the right way to meet the needs of computational biology. The inclusion of prior knowledge and topological assumptions on the underlying network structure reduce the model space complexity. Even then, efficient estimation algorithms are needed to make probabilistic inference possible at all.

Computational Biology

The Gene Center Munich offers the unique opportunity to put theoretical results into practice. In a joint effort with experimentalists, we impose well-defined perturbations on the cell’s transcriptional machinery, then collect large scale protein and mRNA expression data and analyze them with our network models to shed light on the process of transcriptional regulation. Ultimately, we want to gain an inductive understanding of the genotype- phenotype mapping on the cellular level.

We work in close collaboration with the Gene Center's protein bioinformatics group of Johannes Soeding.

Probabilistic Graphical Models for Biological Data

Advances in molecular biology crucially depend on the correct interpretation of vast amounts of experimental data. Since this data is flawed with measurement errors, our goal is to provide the right probabilistic models that lead to relevant and provable conclusions. Our group is working in the field of statistical learning, both theoretically and practically.

Reconstruction of Regulatory Networks
through active Interventions

Nested Effects Models (NEMs) are models which are particularly well suited to the analysis of high dimensional data that has been obtained as response to an experimental intervention into a regulatory system. We participate in the development of the theory of NEMs. We construct stable and efficient algorithms for the estimation of large-scale NEMs. There is a strong need for generalizations of the NEM that permit a more flexible modelling of the regulatory structure. Essentially, we try to be as close as possible to a physical model, without introducing too many degrees of freedom.

Functional Analysis of Protein Complexes
involved in transcriptional Regulation

The mediator complex is a multi protein complex which plays an essential role in the modulation of transcriptional activity. It could be shown that it consists of functionally distinct subcomplexes. We want to identify and separate these subcomplexes and characterise their role in the process of transcriptional regulation. Deletions or modifications of parts of the mediator subcomplex lead to a variety of alterations in the expression profile of a yeast cell. The methodologies developed in our group are applied to gene expression data generated in the Cramer Lab, in order to investigate the roles of the mediator complex components. This happens in close collaboration with the Soeding Group, whose expertise in sequence analysis and protein structure prediction complements our approaches. In another collaborative project with the Foerstemann Group, similar computational techniques will be used to find out how microRNAs orchestrate the transcription of groups of genes.

Best Viewed on Firefox

We strongly recommend using Firefox to fully enjoy this site