Proceeding talk – Theme: Systems.
Abstract
Classical hallmarks of cancer like self-sufficient growth, evading apoptosis and insensitivity to anti-proliferative signals originate from dysregulation of cellular signaling pathways. The Systems Biology approach of establishing cell type-specific mathematical models for understanding the time- and dose-dependencies of cellular interaction networks provides a promising new concept in cancer research for understanding and treating such dysfunctions. One limitation for the reconstruction of differences of the network structures between healthy and malignant cells is the combinatorial complexity which hampers classical model selection techniques. Moreover, parameter estimation is very challenging from the numerical and statistical point of view for ordinary differential equation models (ODEs) because of nonlinearity and limited precision of the solutions and their derivatives. Here, we present an implementation of L1-penalization for identification of cell-type specific alterations in signaling pathways. Robustness and validity of our implementation is proven by using 500 different data setups from the DREAM6 parameter estimation benchmark.
Authors
Bernhard Steiert, University of Freiburg, Germany
Jens Timmer, University of Freiburg, Germany
Clemens Kreutz, Germany University of Freiburg, Germany