Planned analysis
The preliminary results suggest that ‘sampling’ movements of the antenna could lead to output that can differentiate gradient direction without the need for movemnt or bilateral comparisons (Klinotaxis or tropotaxis) . However, the results of this simple model depend on tuning of the time and length scale parameters, as well as the nonlinearity of the time-dependent response. In addition, the weights in the model were set equal to 1, for simplicity, and the output was taken as a linear sum. Current experimental work seeks to take recordings of the input odor levels, ORNs, and the output neuron, such that the model parameters and nonlinearities can be directly informed by the data. This also gives the opportunity to use actual recordings from the cockroach ORNs as input to the model, in order to test hypotheses for neural circuit mechanisms.
The planned analysis will further develop the computational model to test the sensitivity of gradient sensing to different parameters, configurations, and gradient conditions, incorporate neural recordings from experiments, and constrain input-output properties based on the data.
There are many questions raised by the preliminary results, that we wish to address.
How to do the timescales of antennal movement and neural activity affect and constrain the gradient sensing ability?
What are the limits of gradient configurations, in terms of spatial and temporal properties, that can be accurately detected, and how do these compare with typical environmental conditions?
Can nonlinear input-output transformations enhance detection ability, given the constraints of the ‘bottleneck’ architecture? How does noise, in the form of turbulent odor plumes or noisy neural activity, affect detection? What happens if the antenna is damaged?
This project will proceed via a close integration of experimental findings in the computational model. Current anatomical studies are examining the density of ORNs synapses on the output neuron, to ask if connection weights depend on location along the antenna. Behavioral studies with different odor environments till track antennal movement during odor tracking tasks. Electrophysiology experiments yield time-dependent activity of the output neuron (\(v(t)\)), along with the odor concentrations along the antenna (\(s_{i}(t)\)) that elicited this output.
We will constrain and modify model structure by the experimental findings, as well as use measured input-output data as a direct test of information processing mechanisms.
The proposed collaborative theoretical-experimental project will enable us to ascertain what neural processing mechanisms are needed to to distinguish biologically-relevant odor gradients, and to address the evolutionary puzzle of the convergent-divergent anatomical architecture