Modelling of Biological Networks

Our group investigates how biological regulatory networks maintain robust function despite internal and external fluctuations. We tackle this question by analysing measurements from single-cell experiments and systematic perturbation screens. This is performed in close collaboration with experimental partners in order to derive mechanistic and predictive mathematical models of biological processes. Our research focuses on cell-to-cell variability and on quantitative modelling of gene expression responses. We employ deterministic, ODE-based modelling, stochastic simulation algorithms and machine learning approaches. We develop computational tools for improved parameter estimation and model discrimination based on multivariate single-cell datasets, and for network reconstruction from perturbation datasets.

Quantitative description of gene expression

Eukaryotic cells sense and process information in order to respond to environmental changes. While the signalling pathways relaying information from the membrane to the nucleus are well characterised, much less is known about decision making at the level of gene expression responses. One focus of our group is to derive a systems-level understanding of gene regulation which describes: (i) the interplay of signalling pathways and transcription factors in complex gene-regulatory networks; and (ii) how gene expression is coordinately controlled at the transcriptional and post-transcriptional level. Current areas of interest include alternative splicing networks, piRNA-mediated gene silencing and DNA (de)methylation. We tackle these questions by integrating systematic perturbation screens and multi-OMICS data to derive predictive mathematical models.

Quantitative modeling of splicing factor specificity. Binding landscapes of the splicing factor U2AF65 on RNA were determined by in vitro iCLIP experiments, in which transcripts are incubated with increasing concentrations of the recombinant proteins (left). RNA-protein affinities were quantified by fitting a simple binding model to the data (right). This model can be used to study the impact of added regulatory proteins, and to quantitatively compare the in vitro binding behavior with that in living cells. This project is a cooperation with the Koenig group. See Sutandy et al. (2018) for details .


Causes and consequences of cellular heterogeneity

The second focus of our group is the quantitative description of cellular heterogeneity. Even genetically identical cells frequently respond in different ways to the same external stimulus, leading to differences in differentiation programs, drug resistance and viral pathogenesis. Together with experimental partners, we employ live-cell imaging approaches to calibrate stochastic and deterministic models of cell population heterogeneity. In particular, we focus on estrogen- and TGFb-induced signalling events, and investigate how cell cycle networks function robustly despite fluctuations in key regulators. We employ our models to: (i) derive experimentally testable hypotheses about the causes and consequences of cellular heterogeneity; (ii) better understand therapeutic intervention strategies. We are always interested in applications from PhD students, postdocs or students who want to write their Master's thesis in our group.

Cellular heterogeneity in estrogen-induced transcription. Live-cell imaging of nascent transcription at two endogenous GREB1 alleles in the same cell using the PP7/PCP system demonstrates uncorrelated temporal fluctuations of sister alleles (top). Transcriptional bursting can be explained by a stochastic two-state promoter model, in which estrogen regulates the burst frequency and cell-to-cell variability is assumed for the rates of transcript initiation and elongation (bottom). See Fritzsch et al. (2018) for details.