Give


ECE Seminar Series


Joint School of Biomedical Engineering and Electrical and Computer Engineering Department Seminar

Title: Bayesian Inference of Biological Random Walks
Speaker: Jean-Baptiste Masson
Affiliation: Institut Pasteur, Paris, France
Janelia Farm Research Campus, Ashburn, VA

Day: Monday, November 10, 2014
Time: 12:00 pm - 12:50 pm
Location: Behavioral Sciences Building, Room 103

Abstract: Advances in biomolecule tagging allows recording of either dense, short, space filling trajectories, or long, sparse trajectories. While randomly walking in their environments, biomolecules reveal information about the properties and interactions they have with their environments. We are interested in extracting physical properties of complex heterogeneous environments from these trajectories. We demonstrate that by modeling random walks with the overdamped Langevin equation and analyzing the trajectories using Bayesian inference we can infer large scales maps of diffusivities, forces and in some cases interaction potentials. We introduce a user-friendly software, InferenceMAP, allowing easy use of Bayesian inference for single molecule analysis. We discuss statistical modeling of trajectories, multiscale analysis, the consequences of high heterogeneous media and the Ito/Stratonovitch/Hanggi dilemma applied to single molecule environment inference. Finally, we demonstrate that for large number of single molecule experiments, tracking is an unnecessary step and introduce a simple graph theory based algorithm allowing mapping without tracking.