We perform statistical analysis of high-throughput data. For this purpose, we adapt and improve existing methodology and optimize data analysis pipelines for application-specific demands. Comprehensive studies are conducted to assess the performance of available analysis methods and to derive guidelines for selecting best-performing approaches.
Our focus is on stochastic modeling in the life sciences. We cover applications ranging from population genetics to chemical reaction networks, in particular intra-cellular processes. Methods range from stochastic processes to to approaches from machine learning. A future focus will be on human genetic diversity and applications in forensics. The main goal of the group is to provide mathematical models which are rich enough in order to produce insights in applications, but simple enough to be tractable.
Modelling of signalling, metabolic and gene regulatory pathways
We apply differential equations to mathematically model inter- and intracellular networks. The kinetic parameters of the models are determined based on time-resolved experimental data. Experimental design and model reduction strategies are applied to obtain predictive models whose complexity is tailored to the information content of the data. The goal is to obtain insights into the biological systems that can not be obtained by experimental techniques.
Machine learning in medicine / Clinical epidemiology
We develop approaches for modeling complex structures in clinical settings, such as time-to-event data with multiple event types, and connecting these to molecular measurements. This includes machine learning for developing and evaluating prognostic and predictive signatures, and in particular deep learning approaches that allow for statistical inference and provide reliable evidence.