Dresden 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 19: Poster VII

BP 19.27: Poster

Dienstag, 17. März 2020, 14:00–16:00, P2/3OG

Bayesian gradient-sensing in the presence of noise — •Maja Novak and Benjamin M. Friedrich — TU Dresden (CFAED, PoL), Dresden, Germany

Chemotaxis, the navigation of biological cells in external concentration fields, guides foraging bacteria to food patches, immune cells to inflammation sites, or sperm cells to the egg. Chemotaxis strategies must be adapted to sensing and motility noise, inevitable at the microscopic scales of cells, by optimal filtering of chemosensorial input and choice of chemotaxis strategy. A key question is: how to combine most recent and previous sensory input?

We present an information-theoretic framework of optimal gradient-sensing and chemotactic navigation, based on Bayesian sequential estimation. Remarkably, the Bayesian strategy optimally combines "temporal comparison" and "spatial comparison", two distinct gradient sensing strategies employed by biological cells. The width of likelihood estimates of individual agents provides a reliable proxy for the dispersion of direction angles of an ensemble, reflecting the consistency of our approach.

We investigate a search strategy that maximizes the expected information gain in each time step, generalizing the previously proposed "infotaxis" strategy [1] to the case of multiple sensors. We find that agents move slower at locations with low local signal-to-noise ratio to increase the fidelity of gradient measurements.

[1] Vergassola et al. 2007