Stochastic inference of bacterial chemotactic strategies using Kramers-Moyal coefficients
Oliver Pohl (Technische Universitat Berlin)

Bacteria like Escherichia coli and Pseudomonas putida move with alterna- ting runs and reorientations called tumbles that occur with a mean tumble rate.
In the presence of a chemoattractant they move along its gradients thereby performing chemotaxis. We set up a general random-walk model that describes runs and tumbles as a stochastic process of the bacterium's swimming direction and speed. The dynamics includes rotational Brownian motion and shot noise, which initiates sudden tumbling events depending on chemical gradients.
By analyzing experimental data of swimming trajectories, we infer the pa- rameters of our model. For this purpose conditioned moments (also known as Kramers-Moyal coe?cients) are calculated for our shot-noise model and mat- ched to the ones determined from the experimental trajectories. In contrast to common tumbling recognition algorithms no free threshold parameters need to be preset. We show that both bacterial species bias their tumble rates in response to the gradient and quantify this bias. Furthermore, we identify a sub- population of E.coli, which applies a turning angle bias during tumble events when moving in chemical gradients. For P.putida we infer very particular reori- entation statistics and discuss adaptedness to a certain nutrient. Last, we present an extension of our approach, which allows for testing of various chemotactic response functions.