Computational Approaches to Examine Cellular SignallingNetworks

Complex biological systems can only be analysed by utilizing computational and mathematical methods. They are essential for studying the interactions between the components of biological systems and generating an understanding how these interactions give rise to biological functions and mechanisms of cellular signalling networks. In this work, I provide three examples on how the analysis of single cell data derived from live-cell time-lapse microscopy of fluorescent reporter systems benefits from the use of several approaches that originate from computer sciences. The analysis of single cell data faces several challenges ranging from extracting single cell time series from the raw imaging data, identifying signalling classes to the identification of distinct patterns in single cell trajectories. Several methods were introduced in the course of this work to handle the experimental data appropriately. For the extraction of single cell time series, a novel method to track cells in the imaging data based on Coherent Point Drift was introduced. The tracking benefits from the features of the Coherent Point Drift method that correspond to cellular motility patterns. The motion coherence constrain of the Coherent Point Drift mimics the observation that cells do not move independently; they are embedded in a neighbourhood that constrains their freedom of movement. The Dynamic Time Warping framework was established as a useful approach to tackle several issues in the analysis of biological data. The opportunity to quantify the similarity of dynamics granted a new view on single cell data. While calculating the optimal alignment between two single cell trajectories their similarity can be quantified. Dynamic Time Warping was modified so that it constrains the flexibility of the alignment making the alignment more biological relevant. Based on Dynamic Time Warping estimated similarities among individual signalling dynamics distinct signalling classes could be identified in the datasets analysed. Dynamic Time Warping was as well utilized for multivariate time series. This allowed the comparison of single cells while taking the dependence of dynamics of several signalling components within a pathway into account. This gives a new way on the comparison of pathway activity among individual cells. Different signalling pathways exhibit different signalling dynamics. Therefore, two feature detection methods were proposed that aim to quantify signalling dynamics from different angles. The Dynamic Time Warping framework was used to develop a feature detection method that identifies patterns in the time series flexible in the time domain and independent of the scaling. If dynamics lack repetitive patterns dynamics have to be quantified in a different way. Therefore, to identify global dynamics a supervised learning method was developed that reduces the dimensionality of the time series data and identifies fundamental dynamics that compose the observed individual dynamics. To understand how cells, encode the extracellular input and transmit its information to elicit appropriate responses, quantitative time-resolved measurements of pathway activation at the single-cell level was acquired for all three scenarios. The application of the introduced set of tools provided new insight into fundamental biological questions. On the level of the raw imaging data the cell tracking step does only differ slightly between the three biological examples. On the single cell level the three signalling pathways studied exhibit different dynamics and demand therefore different requirements on the analysis. The TGFb pathway is a multi-functional signalling system regulating cellular processes ranging from proliferation and migration to differentiation and cell death. Alterations in the cellular response to TGFb are involved in severe human diseases. It was revealed that the response to a given dose of TGFb is determined cell specifically by the levels of defined signalling proteins and that the observed heterogeneity in signalling protein expression leads to decomposition of cells into classes with qualitatively distinct signalling dynamic. How the dynamics differ among the signalling classes could be quantified using the supervised learning approach. Studies have shown the beneficial effects of hyperthermic treatment during radiation- or chemotherapy of cancers. I aimed to understand how p53 dynamics after genotoxic stress are modulated by temperature across a physiological relevant range. In the range of 33°C to 39°C, pulsatile p53 accumulation dynamics are modulated in frequency. Above 40°C, a temperature that corresponds to mild hyperthermia in the clinical setting, a reversible phase transition towards sustained hyperaccumulation was observed. This disrupts the canonical p53 response to DNA double strand breaks. Above 40°C mild hyperthermia alone was sufficient to induce a p53 response. The view onto the p53 signalling was extended by simultaneously measurement of an additional pathway component. p21 as an inhibitor of cyclin-dependent kinases is the mediator of p53 in growth suppression and a marker of cellular senescence. p53 signalling encodes information about signal intensity, duration and identity in complex dynamics. I studied how these p53 dynamics are related to p21 dynamics in the same cell. It could be shown that p53 and p21 dynamics were not independent and that distinct signalling dynamic shape population response dynamics after application of genotoxic stress. This was shown by using clustering based on multivariate Dynamic Time Warping similarity estimates. Signalling classes and dynamics were connected to cell cycle state.