The group of Hauke Busch focuses on the development and verification of mathematical models for cellular behavior from an initial stimulus to the final phenotype.

In a systems biology approach we combine experimental research on cell-cell communication with the development of appropriate multi-scale dynamic models to investigate the necessary and sufficient control points that lead to cell proliferation, differentiation, migration or death.

We adapt concepts from non-linear dynamics and complex systems to develop appropriate dynamic models unraveling self-organizing properties in cellular behavior. Such behavior in a multicellular environment is most likely the results of time-sequential events, involving protein signaling and gene regulation in feedback-entangled processes lasting several hours.

Systems theory suggests that the slowest evolving variables determine the long term outcome of a system.
In a biological context, it is thus the change in gene expression that reflects the macroscopic decision of a cell.
Formalizing these ideas in a dynamic modeling approach, we will use abstract neural network approaches to reconstruct the dynamic control logic of cellular decision processes based on gene expression kinetics. Time-resolved experimental data will be recorded in our lab under well defined cell culture and context-dependent conditions. Data will be collected on the cell population level using DNA microarrays and RT-PCR as well as on the single cell level by time-lapse microscopy.

Using transcriptome dynamics to unravel NGF-induced differentiation of PC12 cells

The development of mathematical models to simulate cellular behavior is currently hampered by the enormous complexity of biological systems. Building such models thus goes hand in hand with strategies to reduce biological complexity to a computationally and experimentally manageable amount. We propose to reduce this biological complexity by time scale separation of fast protein signaling and slow transcriptome dynamics. Taking NGF-induced differentiation of PC12 cells as a model system, we show how gene expression kinetics can be employed to obtain a global, holistic view on cellular decisions on a time scale of hours to days.

Neurotrophins, exemplified by nerve growth factor (NGF), exert a variety of actions on their targets, including proliferation, neurite growth, neuro-transmission or differentiation. We use PC12 cells as a well-established model to elucidate cellular control points that are necessary and sufficient for NGF-induced differentiation of PC12 cells. Good progress has been made in uncovering initial steps by which NGF acts on receptor-dependent protein signaling pathways. However, it is still unclear, which temporal sequence and/or combination of events are necessary to induce and sustain the differentiation on the gene and protein level over a time span of hours to days. In a systems biology approach we established a data driven gene regulatory network model from cell-wide transcriptome dynamics to predict putative targets controlling PC12 differentiation. Model predictions have uncovered known and novel players that are currently under investigation on the protein and gene level in our laboratory, which are shedding a new light on the complex orchestration of signaling pathways in the differentiation process of PC12 cells.

Furthermore, we follow a multi-timescale modeling approach by first constructing a dynamic gene regulatory network (DGRN) model from time-resolved microarray data, providing cellular regulation on time scales of hours to days. Subsequently, a protein signaling network model simulating the immediate molecular responses upon NGF stimulation on the time scale of minutes to hours links the fast molecular events to the long-term responses in the cell. This combined approach provides an integrated view on the necessary and sufficient events to start and sustain differentiation processes in PC12 cells.
 

GerontoSys

The Gerontosys project aims to improve the understanding of skin aging processes. The project will in an initial step setup a standardized cell database which consist of skin cells samples taken from humans of different ages. Then the cells activities will be inferred using micro-array techniques in order to identify the underlying gene regulatory network. Of particular interest are the differences between these gene regulatory networks from cells of different age.
 

Topology of gene regulatory networks

The gene regulatory network of a living cell controls and regulates cell processes by a manifold of signaling pathways. The determination of their structure and understanding of their dynamics has become a major task nowadays. This challenge is hindered by the sheer complexity in terms of size and the diversity of interactions to be experimentally determined. We investigate here the question how the details of the gene regulatory network matter for signal propagation. Of major interest is the dependence of the signal propagation on network topology, size, and if a detailed knowledge of kinetic interaction rates is required for an appropriate modeling of the gene regulatory network dynamics. Using artificially generated networks, we find that the network topology itself already heavily dominates network dynamics. This promises to greatly simplify proper modeling by theory and relaxes experimental efforts required.

High-throughput transcript profiling methods allow the simultaneous measurement of many parameters within the cell. Conventional interpretation of such data focuses on clustering according to “guilt by association”, thus yielding classes of correlated gene expression profiles. However, it is still an open question how to infer important gene and/or protein regulators from such large-scale, high-dimensional data. Therefore, the identification of key players from genome-wide microarray experiments should be of general interest.

To this end, we mapped the gene expression time series available from the DREAM (Dialogue for Reverse Engineering Assessments and Methods) challenge onto a two-dimensional space using multidimensional scaling (MDS). We found that the low-dimensional embedding strongly reflected the topology of the underlying gene regulatory networks. Firstly, only very few genes are assigned a large radial distance, which we define as the distance from each point to the center of mass in the low-dimensional projection. By analyzing sub-networks extracted from the known E. coli and S. cerevisiae global gene regulation networks, we demonstrate that those outlying genes also tend to have a low degree of network connectivity, and thus being input/output nodes of the networks. Using artificial networks we find that this property stems from hierarchical networks with a power-law type of connectivity distribution, a modular structure that is prevalent in various known biological networks. Using information on protein interaction partners from the BioGRID database, we could confirm this relationship in further microarray time-series, irrespective of the cell type, organism or cell response.

In general, input/output nodes of a cellular network should have a high impact on the development of a phenotype. To test this assumption we applied our approach to microarray time series data on HGF-induced keratinocyte migration. Genes were assigned a relative impact score on migration according to their spatial location after the MDS mapping. Interestingly, we found few genes having a high score, thus acting as putative key players in the decision towards migration. When inhibiting those key regulator gene products, we found a decrease in migratory activity proportional to the MDS-predicted regulatory impact, confirming our results on the correlation between gene network topology and dynamic gene response. Taken together, we conclude that our MDS analysis is applicable to microarray time series in general, leading to rapid key player identification and novel, experimentally testable hypotheses from high-throughput data.

Systems Biology of cellular decisions and cell-cell communication of the skin

Cell-to-cell communication of skin cells is fundamental in the maintenance of tissue homeostasis, growth and survival. This cross-talk is controlled by a variety of molecules, such as cytokines, and the peculiar composition of those signals drives the cell specific behavior, whether proliferation, migration, differentiation or death. Many of those factors are known, but not yet the complexity of its actions on each cell type. Using a Systems Biology approach combining time-ordered double paracrine communication experiments, transcriptome microarray and use of developed multi-scale dynamic models, we intend to uncover regulatory programs present in skin cells, either in health, as keratinocytes and fibroblasts, or in disease, such as melanoma and its related tumor stroma. Understanding the origin, consequence and time-scale of signals we may be able of controlling cellular
decisions.

Discovering functional relationships of proteins by random walks

Determining functional relatedness of proteins from protein protein interaction (PPI) data is crucial for a proper classification of a proteins function. Considering PPIs as binary interactions, they form a network with proteins as vertices and edges between interacting proteins.  We follow the idea that proximity in a distance metric corresponds to functional relatedness of proteins. The key benefit of the proposed method lies in its capability to make predictions for single proteins, while employing the entire network information, i.e. taking advantage of an holistic view.