The Allen Institute for Brain Science developed data-driven models of the mouse primary visual cortex (area V1) for in silico visual physiology studies with arbitrary visual stimuli. The models integrate information from extensive literature review and data from high throughput experimental pipelines. Models were constructed at two levels of granularity, using either biophysically-detailed or Generalized Leaky Integrate and Fire (GLIF) point-neuron models.
The models contain 17 cell classes, represented by >100 unique neuron models, summarized in the figure below. The biophysical and GLIF model networks use exactly the same graph, i.e. each neuron in one network has an exact counterpart in the other, with the same coordinates, presynaptic sources, and postsynaptic targets. The models span a 845 micron radius. For the biophysically detailed model, the “core” (400 micron radius) is composed of compartmental neuron models and is surrounded by an annulus of leaky-integrate-and-fire neurons, to avoid boundary artifacts. Also included is a thalamic cell model representing a lateral geniculate nucleus (LGN) module that provides visual stimulation to V1 neurons in response to arbitrary visual stimuli.
For more information, refer to this pre-print publication by Billeh, et al., Systematic Integration of Structural and Functional Data into Multi-Scale Models of Mouse Primary Visual Cortex.
Overview of the V1 models. Models consist of one excitatory cell class (E) and three inhibitory classes (i; Pvalb, Sst, Htr3a) in each of four cortical layers - L2/3, L4, L5 and L6; L1 has a single Htr3a inhibitory class. Visual stimulation is conveyed by thalamocortical projections (from the LGN).
A representation of the biophysically detailed network is shown on the left (1% of the neurons in the model). For each cell class, the total number of neurons is indicated, with an exemplary dendritic morphological reconstruction.