Kostas Lagogiannis
Thu 18 Feb 2016, 12:45 - 13:45
4.31/33, IF

If you have a question about this talk, please contact: Steph Smith (ssmith32)

How do you locate a source in a large chemical gradient if you are equipped only with a point sensor?
The larvae of Drosophila Melanogaster have been shown to be able to do just that; they spontaneously crawl towards or away from odours even when bilateral sensing is genetically ablated.
This taxis behaviour is thought to consist of several distinct control mechanisms that trigger specific actions, and is even sometimes characterised as decision-making based on short term memory.

We present the alternative, simpler, hypothesis that taxis results from direct ongoing sensory modulation of continuous lateral oscillations of the anterior body, where the sensory organs lie.
Embedding this hypothesis in a discrete-time agent model  gives a characteristic zig-zag motion pattern and larva-like trajectories that can account for recent data showing strong consistencies in the effects of sensory stimulus and learning induced changes on taxis.
A closer examination of the agent model reveals it executes a gradient-ascent line-search algorithm that can operate with a single-point sensor.
Interestingly neural implementations of this algorithm naturally arise by combining the dynamics from existing models of sensory and motor processes. Therefore the proposed algorithm has a plausible neural implementation.
We discuss the potential broader interest that these insights from larva chemotaxis have for research in other fields such as axon-guidance and single cell taxis.