ZURUECK HOCH VOR INHALT SUCHEN

» Back to overview
Proposing Institution

Institute of Computational Biology - Data-driven computational modeling group - Helmholtz Zentrum München
Project Manager

Dr. Daniel Weindl
IIngolstädter Landstraße 1
85764 Neuherberg
Abstract
In the field of experimental biomedicine, current omics technologies generate increasing amounts of data, but the development of data analysis approaches able to fully exploit these data is not keeping pace. Data analysis is often restricted to detecting mere statistical correlations. Given the detailed information available on the systems of interest, mechanistic modeling can provide deeper insights from the same data. However, such approaches are currently limited to smaller models due to the high computational demands.We want to make use of public omics datasets from the Cancer Cell Line Encyclopedia (CCLE) to parameterize large-scale ordinary differential equation (ODE) models of cancer signaling. In the future, the resulting parameterized models can provide the basis for in silico experiments and predictive modeling in the field of drug development and personalized medicine.To this end, we are developing more efficient approaches for parameter estimation of large-scale ODE models. We are borrowing methods from the field of machine learning and applying hybrid stochastic/deterministic global optimization methods for maximum likelihood-based parameter estimation. However, still millions of simulations of the underlying ODE model need to be performed in the process. Due to the complexity of the model, comprising thousands of state variables and parameters, parameter estimation is computationally demanding.A pilot study has been successfully performed on our institute's infrastructure. However, model complexity and amounts of training data will increase further, and thus, computation time will exceed the capabilities of our local systems. Therefore, we are applying for a test account on SuperMUC to perform scaling studies to evaluate and improve the performance of our algorithms on a larger system with the aim of submitting a full project proposal for SuperMUC later on.

Impressum, Conny Wendler