title: Systematic analysis of time resolved high-throughput data using stochastic network inference methods
creator: Bender, Christian
subject: ddc-570
subject: 570 Life sciences
description: Breast Cancer is the most common cancer in women and is characterised by various deregulations in signalling processes, leading to abnormal proliferation, differentiation or apoptosis. Several treatments for breast cancer exist, including the human monoclonal antibody Trastuzumab and the small molecule erlotinib, which both target and inhibit receptors of the ERBB receptor network. However, signalling processes in cancers, especially under drug treatment are not yet completely understood, and methods that learn treatment specific regulation and signalling patterns on a system- wide view from experimental data are needed. One approach is the reconstruction of interaction networks for genes or proteins under external perturbation, and many different algorithms have been proposed in the past. These include Boolean networks, Bayesian Networks, Dynamic Bayesian networks and differential equation systems, all describing the system on a different level of accuracy and complexity. However, if external perturbation is applied, the targets of the perturbations either have to be known, or only the targets of a single perturbation can be learned directly from data in current algorithms. And in general, dependencies of signalling events at different time points should be included into the modelling frameworks, too. This work proposes a novel approach to learn networks from longitudinal and externally perturbed data, called Dynamic Deterministic Effects Propagation Networks (DDEPN )'. Nodes in the network correspond to genes or proteins, selected from a particular biological system, while edges describe the interactions between the nodes. DDEPN models the activity of a node as boolean variable (either active or passive) and creates an activity profile of all nodes for the given time frame, depending on a given network structure. The activity profile is assessed by a likelihood score that describes the probability of the measured data given the activity profile. A network structure that fits best the measured data is identified by modifying the network such that the likelihood score is optimised. DDEPN is applied to a phosphoproteomic dataset from the ERBB signalling cascade, as well as to gene expression data measuring cell cycle related genes. Known signalling cascades from the ERBB and cell cycle networks could be successfully reconstructed and DDEPN also outperformed related network inference approaches. Further, in the ERBB data set, the combined application of the drugs erlotinib and Trastuzumab to the breast cancer cell line HCC1954 resulted in potent inhibition of growth promoting signalling effects, reflected in the down-regulation of the MAPK and AKT signalling pathways. This suggests that this combination therapy could be also a promising option for treatment of breast cancer patients.
date: 2011
type: Dissertation
type: info:eu-repo/semantics/doctoralThesis
type: NonPeerReviewed
format: application/pdf
identifier: https://archiv.ub.uni-heidelberg.de/volltextserverhttps://archiv.ub.uni-heidelberg.de/volltextserver/12082/1/Thesis_Christian_Bender.pdf
identifier: DOI:10.11588/heidok.00012082
identifier: urn:nbn:de:bsz:16-opus-120823
identifier:   Bender, Christian  (2011) Systematic analysis of time resolved high-throughput data using stochastic network inference methods.  [Dissertation]     
relation: https://archiv.ub.uni-heidelberg.de/volltextserver/12082/
rights: info:eu-repo/semantics/openAccess
rights: http://archiv.ub.uni-heidelberg.de/volltextserver/help/license_urhg.html
language: eng