The Allen Institute for Brain Science has an active research program focused on modeling the activity and behavior of the mammalian brain. Below you can access more information on the creation and use of brain-wide, circuit-level and cell-level biophysical models, modeling tools and publications.
Biologically realistic models of the cortical lamina (Layer 4; L4) of mouse primary visual cortex (V1) were constructed at the biophysically detailed and point-neuron levels, to explore mechanisms underlying cortical computations in extensive simulations with a battery of visual stimuli. Learn more about the Layer 4 model.
MCModels is a Python package providing mesoscale connectivity models for the whole adult mouse brain.
Single neuron models
Biophysically-detailed perisomatic-active models are optimized to reproduce the intrinsic firing patterns and action potential properties of individual cells using electrophysiological recordings and morphological reconstructions. Learn more about these models.
Generalized leaky integrate and fire neuron models
Generalized leaky integrate and fire (GLIF) point-neuron models aim to reproduce the spike times of electrophysiological current clamp data collected from mouse and human cortical neurons. Learn more about the GLIF models in the Allen Cell Types Database here.
Tools and applications
The displacement integro-partial differential equation (DiPDE) population density model refers to a simulation platform for numerically solving the time evolution of coupled networks of neuronal populations. Instead of solving the subthreshold dynamics of individual model leaky-integrate-and-fire (LIF) neurons, DiPDE models the voltage distribution of a population of neurons with a single population density equation. In this way, DiPDE can facilitate the fast exploration of mesoscale (population-level) network topologies, where large populations of neurons are treated as homogeneous with random fine-scale connectivity.
The Brain Modeling Toolkit (bmtk) is a python-based software package for building and simulating large-scale models of neuronal circuits. Modelers can build a network once, or use an existing network, and reuse some of the same files describing the network for simulations across a range of different simulators, without having to write converters or adapters. The toolkit provides interfaces to the NEURON simulator for biophysically detailed network, NEST simulator for point-neuron network, DiPDE for populational statistical simulation, LGNModel for filter models, and TensorFlow for convolutional neural networks.