Basic computational models in neuroscience Flashcards
Why make computational models? Isn’t neuroscience complicated enough already?
The goal of computational models is to make our life easier: we can easily test a hypothesis in a system that is way less complicated than the real brain.
How does making computational models make our lives easier
We can model crucial elements to make simple predictions about process. If our hypothesis has a red flag, we’ll see it quickly without having to go all the way in the training/ surgery/ recording/ analysis cycle. They can also help us to spot things that we missed from the data, and refine our experimental design.
Why is learning to model a good idea even if you want to be a ‘pure’ researcher?
Experiment and model are already combined in most cutting-edge research. Even if you plan to be a ‘pure’ experimentalist, knowing about models will make your life much easier when you decide to collaborate with a modeller.
Describe the research Jorge described as a practical example of modelling
A recent experimental study in our group showed, using ferrets, that different (low/high) levels of awareness lead to different oscillatory patterns in several brain areas. Changes in awareness may lead to modulations of visual and parietal cortices from frontal cortex, but technical limitations prevented us to precisely identify, in our experiment, the circuits responsible for these effects.
For this reason, we built a computational model to reproduce the patterns of oscillatory brain activity. Among other things, the model predicted which anatomical connections (from the hundreds of possible options) may produce such activity patterns.
Are computational and theoretical models suited for more molecular process or systems processes?
Computational and theoretical models can be used to describe neural systems at many different scales –from biophysical processes inside synapses to full brain dynamics. They also embrace different approaches, from highly detailed biophysical descriptions to more abstract models.
What could be inferred from differences in the model predictions and the observed data?
Whatever the shape and colour, computational models can use existing knowledge and data to deliver predictions, which then can be used in the design of present and future experiments.
Differences in model predictions could suggest additional mechanisms which are not accounted for in the model. Parameter values and interactions can then be tweaked to become closer to the predicted data.
What else can be derived from models, closely related to predictions?
Models can also be used to make postdictions (Abbott, 2008): to explain known phenomena from a different perspective, or to integrate previous findings into a larger, coherent formal framework. Postdictions can tell you if your assumptions are sufficient
From a more methodological point of view, computational models in neuroscience come from different mathematical branches. Name these (3)
1) The statistical approach (for example, Bayesian neural models)
2) The computer science approach (such as artificial neural networks, deep learning…);while artificial neural networks are not the same, they can behave like real networks
3) The biophysical approach (such as biophysical models of neurons, tools coming from physics and complex systems science)
Which of these three mathematical branches are we most concerned with? What is it most concerned with?
Biophysical; based on what actually happens in the model. The level of detail is up to you
Membrane potential arises from a separation of positive and negative charges across the cell membrane. Where do these excess charges largely stem from?
Excess charges are concentrated directly at the membrane (in- and outside). Each excess is only a tiny fraction ot the total number of ions in/outside the cell
What is the use of having a membrane potential? (high energy demand)
• rapid electrical signalling possible; rapid transport of signals through CNS
• You only have to open “the gates” to use ion gradients to change the voltage
How do you calculate the membrane potential?
Vm= V in– V out
intracellular voltage - extracellular voltage
What is meant by the resting potential and what is the typical voltage?
•Resting membrane potential: “undisturbed” state
•Usually RMP ~ -85 to -60 mV. Thus, excess negative charge inside cell at rest
What is meant by polarisation?
Any potential difference V in– V out that is different from zero. In practice: Vm is negative
What is meant by de-polarisation
loss of (negative) polarization, so Vm moves towards 0 mV
What is meant by hyper-polarisation?
Extra (negative) polarisation, so Vm moves away from 0 mV, becomes more negative than rest
What is meant by the direction of current flow? I.e if ions were moving in and out of the cell, which direction would the current be flowing?
Direction of current flow is defined as direction of net movement of positive charge e.g. “inward current”: net movement of + charge into cell (or: negative charge leaves cell to outside!)
Name and describe the two driving forces at the membrane
Electrical driving force: depends on membrane potential (+ions attracted to inside)
Chemical driving force: depends on concentration difference of ion across membrane