Lecture 27- Modelling of neurons Flashcards
What are models?
- Any description of a process is a model of that process
- A model can be verbal - a description which does not specify a quantitative relationship between elements
- In mathematical models, relationships between elements are described quantitatively – Can be realistic or abstracted
-two distinct forms: verbal (no specification of quantity) it cannot be falsified, the useful ones are mathematical model (there the verbal ones should get) this allows you to test
What is the key point of all models?
- Any model must make testable predictions about the modelled system to be useful
- Without testable predictions computer modelling is worthless
- Biological models (eg mice for humans) must also make testable predictions
- if cannot be tested then not very useful
- models are not imutable, the data must be the bedrock, the models are about predictions
- if want to understand human memory using mice, you are using mice as the model for humans
- must refine models
Why is it good to use computer simulations?
- Uses mathematical models of elements of the nervous system to predict the activity of neurons during a behaviour
- Depends on validity of models and so tests ideas about how neurons or their components function
- Can reveal “emergent properties” of a network
- they don’t have to be accurate (abstract) can add more models in
- once it is setup and the data is in, the validity
- emergent properties are revealed by the models
Why is it good to use computer simulations? (II)
- Even simple behaviours require the coordinated activity of thousands of neurons
- At single cell level, behaviour is the result of interactions of many different mechanisms
- Typical biological models use stereotyped stimuli and do not reveal the full response patterns involved in a behaviour
- In many cases, the interactions between specific neurons controlling a behaviour are too difficult to determine
-to specify each neurons individually hat is impossible at the moment, so computers can with models do this -have to validate the result of computer to brain
What are the emergent properties?
- Properties of a system that cannot be predicted simply from the properties of its individual elements
- In the nervous system, content addressable memory is an emergent property of networks obeying Hebb’s rule
-subset of neurons synchronize= one of the emergent properties in the ENS, that is what they found, only shown up once run as a network in a model= similar happens in the brain (alpha rhythm etc.)
What are compartmental models?
- Incorporate mathematical descriptions of the ion channels in neuronal membranes – And movements of ions in the cytoplasm
- Also include models of spatial relationships between different components of the neuron
- Very good for identifying what a single neuron might do and how specific channels or currents fit in to behaviour
- Have very heavy computational requirements and so are often not used for large networks
- this is what was used to work out how the AIS works
- heavy computational requirements to get useful information,
What is the compartmental model of an ISN?
- myeteric neuron
- one is the model output and one is the intracellular recording
- the right is the model -the left is the real one
What types of ion channels were included in the compartmental model of an ISN?
- Leak conductance – resting K+ and Cl-
- Voltage dependent Na+ - NaV1.3, NaV1.6, NaV1.9 • Voltage dependent K+ - generic
- Voltage dependent Ca2+ - N type
- Calcium dependent K+ - BK (also voltage- dependent) & IK
- Calcium dependent cation conductance
- HCN channel – produces current known as Ih
- Also modelled intracellular Ca2+ handling
- the things built into the model
- these channels have been shown to be there, but how they interact is important as they change when inflammation and keep the change after
- can vary the channels in the model as opposed to in the body which difficult
What is true of the different channels?
-different channels, different outcomes
What is the example from the ENS?
- There is a network of interconnected sensory neurons
- Descending (anal) interneurons come in several classes with different connections
- There is only one class of ascending (oral) interneuron
– Guinea-pig small intestine
What are the two way to go once have a network?
• Realistic models
– All neurons modelled with mathematical precision covering connections, transmitter release, ion channels and intracellular molecular dynamics
– Very heavy computational requirements for mammalian circuits
• Abstracted models
– Assumptions made about specific components to obtain general rules for the system
– Can vary from completely top-down to various hybrids of realism and abstraction
-under normal circumstances you have to abstract, have to assume about how the networks work, how they interact
What are the simple integrate-and-fire neurons?
- Basis of most neural network models that “simulate” cognitive processes
- Basis of most neural network models that try to simulate cognitive processes – Elementary abstraction of action potential
- Each neuron adds its synaptic inputs and has a firing rule to generate stereotypes action potentials
- Classic version is the McCulloch-Pitts neuron (originally an electronic model)
- this is what is used in models,= McCulloch PItts model= electrical circuit etc.
- voice recognition systems, automatic car parking,
What is this?
- highly abstracted model of the INS
- properties of sensory neurons could be related to patterns that produce segmentation in the ENS, interactions of all sorts of things in the system
- model implemented in a mathematical model
- can be used to explore key problems -clumper model (all excitatory neurons are lumped together)
What is the starting functional data?
- Decanoicacidin lumen of guinea-pig jejunum
- TRAM34 and clotrimazole block AHPs (IK) in intrinsic sensory neurons
- Periods of activity increase but periods of quiescence do not change
- block the after deploarisatoon in ENS and what happens when blcoked, burst of activity quiescent activity
- blocked the hyperpolarising potentials= get huge increase in activity, so system is more excitable
- strange result= the activity periods increased but quiescent periods stayed the same= surprise
How can you use model outputs?
– neuron and muscle activity
- build model and see what happens when change
- much longer excitatory periods but inhibitory remain the same= changing the hyperpolarising the potential= matched the model though
