[seminar] Seminar Novak, 7. 2., 11h, F-201

["Dario Jukić"]

Dear colleagues,

I am happy to announce joint QuantiXLie and Croatian Biophysical Society
seminar that will be given by Maja Novak from Dresden.

Title:
Sequential Bayesian estimation for gradient sensing with spatial comparison

Time and place:
Thursday, 7. 2. 2019., 11:00h,
Department of Physics (F-201).

Abstract:
Sensing by spatial comparison is an important mechanism that enables large
and slow cells to efficiently navigate towards the maximum of the
concentration field. Even the sensing by temporal comparison, like in
navigation of sperm cells in open sea, can be mapped to the sensing by
spatial comparison. It is known that molecular shot noise, as well as the
motility noise, limit the accuracy of concentration and gradient sensing.
The framework of sequential Bayesian estimation provides the efficient way
to improve this accuracy. Here we study the minimal model of 2D agent that
measures stationary gradient and concentration with the presence of
sensing and motility noise and updates its likelihood function of the
environment. We consider two modes of Bayesian estimation, depending on
the environment model with and without rotational diffusion, which agent
uses to predict the likelihood function. We show that model with
rotational diffusion is superior, with the degree of improvement depending
on the signal-to-noise ratio of the environment. We also show that the
width and information present in the likelihood function of the
environment depend solely on the signal-to-noise ratio. Our study reveals
non-intuitive result for the width of likelihood function when motility
noise is present. In conclusion, we propose the implementation of the
sequential Bayesian estimation in single cells, as an iterative scheme
that mimics the memory on the environmental state.

Regards,
Dario Jukic