Heinz Koeppl: "Statistical inference of cellular behavior from data"
von 11:15 bis 12:00
|Wo||FRIAS Seminar Room, Albertstr. 19, 79104 Freiburg|
|Kontakttelefon||+49 761 203 97418|
Open to University members
BISON Group - Biomolecular Signaling & Control Automatic Control Lab, ETH Zurich, Switzerland
Statistical inference of cellular behavior from data
The computational reconstruction of molecular processes constituting cellular behavior from experimental data is at the very heart of systems biology. It is a challenging problem because of the large uncertainty in the data, our limited access to the underlying process and our incomplete understanding of the process itself. Among data types we may discriminate between high-dimensional data with usually poor temporal resolution and accuracy and data of low dimension but good temporal resolution that is usually available at the single-cell level. In this talk, I will provide an overview of the activities within my group to address reconstruction problems related to the latter data type. I will discuss novel statistical state and parameter inference methods for stochastic kinetic models of bio-molecular processes involving compute-intensive Markov chain Monte Carlo and Sequential Monte Carlo algorithms. More specifically, I will show how mixed-effect models from statistics can be used to properly account for the significant cell-to-cell variability present in a clonal cell population. Inference techniques applicable to time-lapsed fluorescence microscopy data and to flow cytometry data will be discussed. I will also briefly touch upon the issue of optimal experiment design for the purpose of reconstruction.