The first version of the workflow can be found in the Brain Simulation Platform.
If you do not have an account to access the Brain Simulation Platform, please email email@example.com
- Lindroos R, Dorst MC, Du K, Filipovic M, Keller D, . . . Hellgren Kotaleski J (2018). Basal Ganglia Neuromodulation Over Multiple Temporal and Structural Scales-Simulations of Direct Pathway MSNs Investigate the Fast Onset of Dopaminergic Effects and Predict the Role of Kv4.2. Frontiers in Neural Circuits.
- Brocke E, Bhalla US, Djurfeldt M, Hellgren Kotaleski J & Hanke M (2016). Efficient Integration of Coupled Electrical-Chemical Systems in Multiscale Neuronal Simulations. Frontiers in Computational Neuroscience.
Brocke, E., Djurfeldt, M., Bhalla, U.S, Hellgren Kotaleski, J., Hanke M. (2017) Multirate method for co-simulation of electrical-chemical systems in multiscale modeling. J Comput Neurosci 42, 245–256 (2017).
Multiscale models are models that expand and link beyond the domain of single-level models. There are many possible levels of multiscale modelling relevant to brain simulation, for example: systems models, point-neuron models, morphologically-detailed models, different-equation subcellular models and molecular dynamics models.
What makes multiscale models important?
Multiscale models allow us to understand how changes occurring at one level of simulation can propagate to higher levels.
What are the specific questions we want to address in the HBP?
We aim to answer two types of questions:
1. How can parameters be propagated across scales?
2. How can simulations at multiple levels be run concurrently?
In order to propagate parameters across scales, the equations used at one scale, must be compatible with those used at another. For example, if one examines the rate of spine head clearance of calcium, for a model in which this was done by kinetic models of calcium pumps and exchangers, one could measure a calcium clearance time constant and then use this in a model in which calcium decay is represented as an exponentially decaying function.
In order to couple multiple simulations, information between simulators must be exchanged at periodic time intervals. It may be desirable to only model a small portion of the entire system at high resolution due to computational requirements.
What is our specific take?
Our initial multiscale models have focused on integrating intracellular signalling and electrophysiology. The flux of some ions through channels in the cell membrane does not only change the electrical properties of the membrane, e.g. the membrane potential, but can also influence intracellular cascades, that ultimately change the properties of the channels themselves. These two functions of single ions are typically modelled using different types of models. The electrical properties are modelled in one type of model, while the intracellular changes are modelled using another. When combining scales like these it is important to closely control the numerical stability as this can be compromised in the multiscale aproach (Brocke et al., 2016, 2017).
Thus, we have started out by simply adding more detail to simulations running at a coarser scale. When computational power becomes limiting we will explore other methods of multiscale simulation.
We are currently identifying what is needed and how it can be achieved. We are also looking into possible limitations. A first workflow on how to transform intracellular cascades into NEURON readable mod files has been implemented as a Jupyter notebook. This is available in the Brain Simulation Platform’s Subcellular Modelling Collab.
As a proof of concept, this workflow has also been used to transform, insert and simulate a dopamine induced intracellular signalling cascade into an electrophysiological model of a single neuron (Lindroos et al., 2018). Specifically, this technique enables investigation of time scales involved in for example neuromodulation. A simplified version of this kinetic scheme was recently used in the striatal network model (Hjort et al., 2020, see the Basal Ganglia tab). During the next phase of HBP this collaborative approach will continue as the neuromodulatory system in the network is further developed and explored.
Further, in collaboration with the signalling cascade group a more versatile and general tool for conversion is under construction.
Since this work is in its initial phase, a small team is working to address these questions:
KTH Royal Institute of Technology, Sweden
Robert Lindroos, Karolinska Institutet, Sweden
KTH Royal Institute of Technology, Sweden
Daniel Keller, Blue Brain Project, École Polytechnique Fédérale de Lausanne, Switzerland
Benefits to the Community
Model use: you can use the workflow from within the Brain Simulation Platform’s Subcellular Modelling Collab.
Participate in community modelling: By participating you can expose your models to a wider community. We would be happy to hear from you if you would like to get involved and contribute to our community effort. You can also send a request on what type of models you think are needed.
Please contact one of the team above, if you would like to participate in the effort.