From perception to behaviour
Initially scaled a drosophila adult spiking network (Wessnitzer et al. 2012) to larva size by collecting known data on the number of ORNs, projection neurons and Kenyon cells. Verified it can display conditioning with a simple spike-timing dependent plasticity rule. This network model inspired simplification to formal based network, removing the complexity of the spike-timings to consider conditioning under simple network architectures and learning rules with addition of habituation in the antennal lobe. However, it was soon apparent that to understand how learning alters behaviour it is necessary to first look at how perception drives chemotaxis by examining initial sensory coding and chemotaxis algorithms. Once this level is understood we can then make progress into understanding how conditi0ning modifies preferences for particular odours and embed the simplified learning circuit to a chemotactic agent simulation.
Building apparatus for automating larva experiments.
The idea is to use robotics to build a machine that is able to pick and place larva from vials to petridish for tracking and further transfer them between petridishes. At this stage we have come up with a tool design that is able to pick-up and drop larva. This is next to be mounted on a linear manipulator in order to test the efficiency of the prototype in successfully transfering larva on a petridish through automated control. In parallel, we are building a FIM table (http://fim.uni-muenster.de/) and integrate it to the prototype machine.
An agent-based model of larval chemotaxis has been developed, allowing us to test how well current hypothesised larval chemotactic mechanisms can account for larval behaviour in odour environments.
The video below demonstrates some variants of our model larvae navigating in an ‘choice-assay’-like environment.
An unbiased behavioral classifier has been developed. The pipeline combines the ‘eigenmaggot’ representation of behavior with model based clustering. Recent development is that locomotion speed has been incorporated to the framework.
Virtual larval brain
We are experiencing success aligning the larval CNS and brain Light Microscopy image stacks to a template, with the aim of integrating this into the Virtual Fly Brain website. We aim to have a prototype available in the new year.
Modelling the larval body
A simplified mathematical model of the larval body has been constructed. Control theoretic techniques are being applied in order to better understand the active and passive dynamics of larval locomotion. This approach will allow us to interpret recent results from the ventral nerve cord within the context of embodied sensorimotor control, and may suggest novel design principles for soft robotic systems.
Examining the activity of neurones in Drosophila larvae
A luminescent approach has been used to record the activity of specific neurones in Drosophila larvae. Currently, both genetic and pharmaceutical approaches are used to test this system.