We use theoretical and computational approaches to study regulatory and cellular networks involved in cell-cycle regulation, oncogenic signaling and molecular oscillators. We combine functional data sources such as microarrays, chromatin-immunoprecipitation (ChIP) and genomes to infer regulatory dependencies between genes and regulatory proteins involved in cell proliferation. Well-characterized networks such as the cell cycle in yeast offer opportunities to study how information propagates through these systems and identify links which are crucial in mediating function, or alternatively that confer robustness properties. One thematic focus is the study of biomolecular oscillators, in particular the circadian clock which drives periodic behavior and physiology in most living organisms. Recently we studied how light and the core oscillators coregulate genes involved in specific output circuits in the fly. We are also interested in the collective properties in populations of cellular oscillators, such as the emergence of synchronization.

Our goal is to study the role of cell-cell communication during metastatic colonization in mouse liver using a xenograft model. In collaboration with Joerg Huelsken we successfully developed an original transcription profiling technique for chimeric mRNA samples. This will now be applied to study metastasis progression in mouse liver with tumor and stroma specific resolution without the need for delicate microdissections.

This project consists of two parts, one in which we developed a signal estimation technique for tiling arrays, in particular the detection of binding sites in genome-wide chromatin immuno-precipitation experiments (ChIP). Secondly, we studied the functional link between protein‑DNA interaction networks and gene regulation by quantifying how the expression of genes that harbor c‑Myc or Sp1 sites responds across a large collection of tissues. Using regression models, we found that genes with both c‑Myc and Sp1 binding sites have a distinct expression signature as compared to genes with either site alone. We showed that this property characterizes c‑Myc regulation much more broadly and identified combinatorial functions that explain expression of human target genes.

We show how dynamical modeling can be applied in the yeast cell cycle to make predictions about functional interactions that are crucial for S phase and DNA damage checkpoint efficiencies. Our work studies checkpoint responses in silico in the budding yeast by applying an information theoretic analysis to a stochastic version of the cell-cycle model by Li et al. Specifically, we quantify the responses to activated S‑phase and DNA damage checkpoints as a function of structural network perturbations. Our approach permits to identify crucial network components for optimal checkpoint efficiencies in response to environmental fluctuations. We evolve networks by adding links that successively improve checkpoint function. Intriguingly, this analysis points to the SBF and MBF complexes as playing a key role for optimality, and connects with recent discussions on the redundancy and target specificity of these regulators.

Cell‑autonomous and self‑sustained molecular oscillators orchestrate circadian behavior and physiology in mammals. Such rhythms were recently measured both in individual cells and populations using bioluminescence reporters. Using a combination of experiment and modeling we could show that the dominant cause for amplitude reduction was desynchronization of self-sustained oscillators rather than decay of individual oscillators. Recently we studied a natural extension of the static frequencies with two biologically relevant additions: (i) the recent observations by Carr and Whitmore that frequencies drift in time, (ii) the possibility that oscillators are coupled, for example through gap-junctions or inter-cellular signaling by diffusible secreted molecules. Besides the frequency dispersion s _ƒ the model introduces a new, previously uncharacterized parameter g: an inverse correlation time describing the frequency drifts. We also introduced coupling among the oscillators to examine the possibility of collective synchronization in oscillator populations. Using analytical methods from stochastic physics we could derive a formula for the critical coupling strength in relation to the other model parameters g and s _ƒ . Ongoing work now applies these results to estimate frequency drifts and inter-cellular coupling from real data (Fig. 1).

Figure 1: Analysis of circadian cell-culture bioluminescence data. Left: Raw data reproduced from Nagoshi et al.. Inset: logarithmic scale emphasizes the exponential signal decrease reflecting cell death with half life 3.22 days. B) Luciferase reporter construct in the Bmal locus. Right: Detrended data with the modeled envelope (red) superimposed. Residual oscillations at long times (experimental amplitudes are larger than the predicted red curve for t >12 days) are expected due to the finite number of cells left in the population.

Keywords

Bioinformatics, systems biology, cancer genomics, gene expression, biological oscillators