Circuit models for direction selectivity: V1 DS Simple Cells

Objectives

To learn about population models in the WM framework.
To use the population GUI to explore model architecture.
Understand how to implement direction selectivity in a biologically plausible network.

References:

Baker PM, Bair W (2012) Inter-neuronal correlation distinguishes mechanisms of direction selectivity in cortical circuit models. J Neurosci 32:8800--8816.

The model used here has documentation on www.iModel.org: DS_Simp
Further documentation for population models in general is available here: WM pop

Network population models like the one used here have considerably more parameters than the filter models of the earlier exercises. The links above have additional information on the parameters that can be specified in the population models.
Examine the contents of the model parameter file
Open the model .moo file from the model home page linked above. Note there are several populations:

LGN and LGN_D are two LGN populations that differ only in response latency (the temporal filter for LGN_D units is delayed by 20 ms relative to LGN).

IN and EX are inhibitory and excitatory V1 units respectively that take input from the LGN population. IN_D and EX_D are V1 units that take input from LGN_D.

DS units take input from all 4 IN and EX populations. The phase of the templates that select the inputs differs between each of the four populations, generating direction selective responses.

Visualize the model with the GUI (graphical user interface).
Open the model viewer from the DS_Simp model home page linked above (the 'View' button). A window with two panels will pop up. Each panel displays a population in the model. At the bottom of each panel is a dropdown menu with a list of the different neuronal populations defined in the model .moo file. You can change which population is displayed in either panel by selecting one of the populations from the dropdown menus.
Click on an EX unit in the left-hand panel and select the LGN population for the right-hand panel. At the top of the window will be displayed the number of connections to the selected unit from the LGN, the total synaptic weight of these connections, and the preferred orientation and spatial RF phase of this unit. Left-clicking on a unit reveals its pre-synaptic connections from the population in the right panel. If you right-click on a unit, it will show projections (post-synaptic connections) in the other panel. If there are no connections between the selected unit and the population in the other panel, no units will be highlighted.
Another feature is the ability to view indirect connections. Select the DS population on the left and the LGN population on the right. Press 'i' on the keyboard - you will have to option to select an intermediate population that receives input from the LGN units and provides input to the DS population. Select the EX population, and in the main window middle click on a DS unit. In the right panel you should see the aggregate LGN input coming to the DS unit through its EX inputs. Take a look at the indirect connections from the LGN and LGN_D units to the DS population. What do you notice about the spatial arrangements of the different inputs?
Run the model using the direction tuning stimulus.
Take a look at the direction tuning of spike outputs from the 4 DS units. Go to your iModel account and run a direction tuning curve with this model. The model to select is DS_Simp. The direction series stimulus file to use is sine_dir.stm. The response file for spikes from the DS units in this model is ds_simp_sp.rsp. Run 5 repeats of the stimuli for 512 time steps, with a grating spatial frequency of 1.2 cyc/deg and temporal frequency of 10 Hz:
  wm mod DS_Simp.moo sine_dir.stm ds_simp_sp.rsp tn 512 sf 1.2 tf 10 stim_nrpt 5
To look at the associated excitatory and inhibitory conductances to a single DS unit run another simulation with the same parameters, but with the response file ds_simp_cond.rsp:
  wm mod DS_Simp.moo sine_dir.stm ds_simp_cond.rsp tn 512 sf 1.2 tf 10 stim_nrpt 5
View the response using the nData browser (nd)
Once the simulations are complete, open one of the nData files. Select the Tuning Curve analysis and set the contrast to 1.0. Verify that all the units are direction selective.
Open the file containing the input conductance traces. Observe how excitatory and inhibitory conductances vary with direction. Note how the time course of the conductances changes as you change direction.
Varying the phase of inputs to the DS cells
This model was inspired by the experimental work of Priebe and Ferster (2005) Direction selectivity of excitation and inhibition in simple cells of the cat primary visual cortex. Neuron 45:133-145. They proposed that direction selectivity is generated by the precise temporal arrangement of excitatory and inhibitory input to DS simple cells.
In the model this can be tested explicitly by varying the phase relationship between excitatory and inhibitory input to the DS cells, and/or the phase relationship between delayed (EX_D, IN_D) and non-delayed (EX, IN) inputs. This is determined in the model in the input objects to the DS units, and can be varied by changing the phase parameter. Open the .moo model parameter file again and scroll down until you find the following lines (in the DS population description):
      ### Template params, ori taken from orimap
      template Gabor # Use this Gabor template, not 'self_lgn'
      sf       2.0   # (cyc/deg) Spatial frequency
      sd_orth  0.16  # (deg) Gaussian SD for Gabor RF, orthog to ori
      sd_par   0.28  # (deg) Gaussian SD for Gabor RF, parallel to ori
      phase    0     # (deg) phase
This appears 4 times in the .moo file, once for each input to the DS units coming from the EX, EX_D, IN and IN_D populations. Note that for each of these inputs, the phase parameter differs. The delayed and non-delayed inputs are set to be 90 degrees apart, while the inhibitory and excitatory inputs are set to be 180 degrees apart.
Try varying the phase difference between these inputs. You can do this by downloading the .moo file, editing the file, and re-uploading it to your account, or by changing the phase parameters directly as command line parameters. For example, to use a template with phase 180 degrees for the EX population input, the syntax to set the parameter value would be pop.ds/input.ex/distrib/phase 180) What is the effect on direction tuning of spikes? How do the conductances to the DS units change? What conclusions can you draw about the importance of the phase relationships between excitation and inhibition and delayed vs. non-delayed input in generating direction selectivity in these simple cells?