¡GoConqr estrena diseño! Cuéntanos qué te parece.

Regístrate gratis

Fichas

Computational Neuroscience

Anna Hogarth Techniques in Neuroscience Year 2 Inglés 97 fichas

Computational Neuroscience

Fichas del conjunto (97)

  1. 1All models are wrong, but some are usefulX
  2. 2What are the two general types of models?Abstract and realistic
  3. 3What was one of the shortcomings of the hodgkin huxley model? What does this not not take into account?Used a wire in the squid axon rather than allowing current to travel through the cytoplasm. Different parts of neurons have different voltages etc.
  4. 4What model was developed to account for what hodgkin and huxley didn't account for?Multi-compartmental models
  5. 5How do models get more realistic?Looking at individual channels, molecules and more and more complex.
  6. 6What happens as models get more abstract?Become simpler
  7. 7What are the six models from the abstract end of the scale to the realistic?1) Binary (McCullogh-Pitts) - Is the neuron on or off
    2) Firing Rate
    3) Integrate and fire
    4) Hodgkin Huxley
    5) Multi-compartmental models
    6) Molecular models
  8. 8What are the pros and cons of abstract to realistic models?1) A: Simple, R: Detailed
    2) A: Hard to relate to biology, R: Contains stuff you can measure/use in lab
    3) A: Few parameters, R: Lots of parameters - every choice requires numbers obtained from experimental data
    4) A: Fast simulation, R: Slow simulation (even with computing power today modelling a full neuron at the molecular level would take a long time)
    5) A: Mathematical analysis (can use algebra to solve more general problems), R: Intractable
    6) A: Generic, R: Specific
  9. 9—Realistic models are far more specific and can't be applied in a general way - ie granule cell isn't the same as the pyramidal cell. It is in a general high level input output way, but at detailed level are very different.Realistic models are far more specific and can't be applied in a general way - ie granule cell isn't the same as the pyramidal cell. It i...
  10. 10Which type of model is better to use?Depends on question - more detailed doesn't always answer the question better.
  11. 11—Top = real neuron recording
    Bottom = Computer model - fairly accurate simulation.    Top = real neuron recording Bottom = Computer model - fairly accurate simulation.
  12. 12What does this show?Many APs overlaid - traces are quite different until they hit threshold and then become quite uniform. Explained by Hodgkin Huxley model and opening of Na+ channels.What does this show?
  13. 13What is responsible for the stereotyped AP? What is the conclusion of this?1) Na+ and K+ channels behave in a very stereotyped and consistent manner.
    2) Height/weight of the spike contains no information - frequency of spikes contains info.
  14. 14How do integrate and fire neuron models work?Remove spike (assumed to be uniform) and and add a fake spike. Model the the return to membrane voltage (the integrate). Direct voltage reset.
  15. 15Where does the leak bucket analogy fail?Goes to zero (which neurons can't)
  16. 16—Dashed line = steady state current - indicates presence of threshold value. Fake spikes inserted when current hits -55mV and then voltage is resetDashed line = steady state current - indicates presence of threshold value. Fake spikes inserted when current hits -55mV and then voltage...
  17. 17—Rate of firing with respect to current input. Goes from 0 to 20Hz very rapidly - same as in real neurons.Rate of firing with respect to current input. Goes from 0 to 20Hz very rapidly - same as in real neurons.
  18. 18What is a difference between integrate and fire model and real neurons?Current input changes with time depending on how many channels are on/off in real neurons. In model input current is fixed with time.
  19. 19How fast can voltage track changes in synaptic input?Depends on membrane capacitance C and and input resistance - wrap into constant t (tau). The smaller tau the faster the neuron and the more quickly it responds to changes in input.
  20. 20What type of changes are real neurons better at tracking?Slow changes
  21. 21—Recording angle of tweaking and voltages of two cells/neurons. Technically very difficult.Recording angle of tweaking and voltages of two cells/neurons. Technically very difficult.
  22. 22—Information required - numbers from experiments, how likely are neurons to be connected to each other etc. Each dot corresponds to a neuron being activated - can see patterns were many are activated. ADVANTAGE: Can look at many neurons at one time. Can then compare model to experiments to see if you can use it to predict outcomes of future experiments.Information required - numbers from experiments, how likely are neurons to be connected to each other etc. Each dot corresponds to a neur...
  23. 23QUESTIONS FOR HANDOUT.X
  24. 24What is modeling a tool for? (2)1) Testing out ideas about the functional significance of an experimental discovery
    2) Language for phrasing ideas
  25. 25Why might it be useful to model how hippocampal fast oscillations might occur through gap junctions?Gap junctions are very difficult to block experimentally - use computer to try and simulate a network of neuron axon-to-axon gap junctions to see if the network supports very fast oscillations.
  26. 26What is the trade off which needs to be made?1) A very accurate and detailed model which may be slow to simulate and hard contain lots of parameters which are hard to fix using experimental data
    and
    2) Very simple models that are easier to simulate and contain few parameters but are inaccurate and may not incorporate important aspects of dynamics
  27. 27What is the simplest model of neurons and their voltage dynamics compare to?Dynamics of a leaky bucket
  28. 28—Where i(t) = rate of water input
    h(t) = rate of change of water height
    l(t) = rate of leak of water    Where i(t) = rate of water input h(t) = rate of change of water height l(t) = rate of leak of water
  29. 29Describe the relationship between h & l - what equation denotes this?The rate of leak i(t) is dependent on the rate of change of water height h(t) - the greater the value of h the greater greater the pressure at the bottom and hence the faster the water leaks out of the bucket.
    l(t) = Gh(t)
    where G is a constant encoding many factors including size of hole, viscosity of water etc..
  30. 30What is the equation for the total amount of change of volume of water?i(t) - Gh(t)
    -- The total rate of change is equal to the rate of water input minus the rate of change of height of water. If rate of input is less than the rate of change of height then then the overall rate of change will be negative (ie the amount of water will be decreasing)
  31. 31What equation describes the volume of water in the bucket?Ch(t) where C is the cross-sectional area of the bucket.
  32. 32How is the rate of change of volume denoted? What is it equal to?1) dCh(t)/dt (eg dx/dt)
    2) dCh(t)/dt = i(t) - Gh(t)
    Rate of change of volume is equal to the total rate of change due to change in input - change in height
  33. 33How can the equation which describes the rate of change of volume be rearranged?dCh(t)/dt= i(t) - Gh(t)
    Ct.(dh/dt) = t(i-Gh)
    dh/dt = 1/C.(i-Gh)
    The rate of change in volume is equal to one over a constant times (input - rate of change of height)
  34. 34Summarise the equation which describes the rate of change of height in words.The rate that the height of the water changes is proportional to the amount of water flowing in minus the amount of water flowing out. (Gh=L)
  35. 35—XX
  36. 36What is the solution to the dh/dt equation?Where T = C/GWhat is the solution to the dh/dt equation?
  37. 37How does this solution make sense with respect to the dh/dt equation? At these conditions what will the height of the water be equal to?When h = i/G (ie Gh=i, change in input =change in height) then in the rate of change equation dh/dt=0
    dh/dt = 1/C(i-Gh)
    if h = i/G
    then: dh/dt = 1/C(i-G.i/G)
    = 1/C(i-i)
    = 0
    The height of the water at this point as described by the h(t) equation will be equal to i/G and constant ie the input over the external factors.
  38. 38How do the dynamics of the water flow change with time?1) The more water in the bucket the greater the pressure and the faster the water will flow out of the bucket
    2) With time the rate of water input matches the output - equilibrium point.
    3) If there is more water input than required for equilibrium the water will pour out faster
    4) If there is less water input than required for equilibrium then the water will pour out more slowly
    5) In either condition the rate of water output moves towards equilibrium ie dh/dt=0
  39. 39—Dynamics of rate of change - the higher the starting input (h(0)) (ie height of water at time 0) the faster the rate of change of height. ie h(0) = 3 has a greater dh/dt than h(0)=2. At h(0) = 0 there is no water in the bucket so the rate of output increases with time as the bucket starts to fill due to input (i). In all scenarios the output moves towards equilibrium. The closer the rate of flow gets towards equilibrium (ie h=i/G or Gh=i) the slower the rate of change -- dh/dt moves towards 0.Dynamics of rate of change - the higher the starting input (h(0)) (ie height of water at time 0) the faster the rate of change of height....
  40. 40How does variable input (rather than constant) effect the solution of the equation?Solution follows the input with a lag set by the constant T (C/G remember from h(t) equation):
    If T is large then the lag is large and the output is 'smoothed' out. If T is small then the lag is smaller and the output follows the input more faithfully. Thinking about neurons - the constant T will define how a neuron responds to a given type of input/firing rate.
  41. 41—XX
  42. 42—XX
  43. 43What are the issues with solving dh(t)/dt = 1/C.(i-Gh)?Often can't solve analytically ie
    1) For many values of i(t)
    2) i(t) is only known numerically
  44. 44What is the simplest approach to solving dh(t)/dt = 1/C.(i-Gh)? Describe a scenario where you might use this.Euler's method:
    You know the value of h(t) but want to approximate h(t+δt) ie the height of water at time t + the height of water over a given increment in time. δt is a small increment in time. δ in this scenario is delta, not the d of dx/dt
  45. 45What does Euler's method use? How are we able to do this?h(t + δt) ≈ h(t) + δt.(dh(t)/dt)
    ie Height of water (h(t)) + height of water over a given increment in time (hδt) is approximately equal to the height of water (h(t)) + the rate of change of water (dh(t)/dt) X that given value of time (δt).
  46. 46If starting at a time of t0 and we know h(t0) what are hn and tn equal to?If starting at a time of t0 and we know h(t0) what are hn and tn equal to?
  47. 47What is hn+1 equal to? What are the possible ways to solve this and the advantages of each?1) Integration on a computer - easy but less sophisticated
    2) Runge-Kutta
    3) Backwards Euler
    Both these methods are more complex and computational costly at each steps but are more accurate.    What is hn+1 equal to? What are the possible ways to solve this and the advantages of each?
  48. 48What creates the voltage difference across a neuron membrane?Ion pumps.
  49. 49How many ions does the Na+/K+-ATPase ion pump move?Sodium-potassium adenosine triphosphatase pump. Three Na+ ions out and two K+ in.
  50. 50What is the typical RMP?EL=-70mV
  51. 51How do excitatory/inhibitory synapses change voltage?E: Increase voltage
    I: Decrease voltage
  52. 52What is a usual threshold voltage? What happens if this is crossed?-55mV - nonlinear cascade producing an AP
  53. 53How wide is an average voltage spike? What value does is reset to?1) 1-2ms
    2) -65mV
  54. 54How long does the refractory period last for approximately?5ms
  55. 55How do capacitance and voltage of a neuron relate to the bucket equation?1) Capacitance (Cm) replaces C - constant which for volume of bucket
    2) Voltage (V) replaces h - height of water
    CmV = the amount of electrical charge which can be stored by the cell replaces volume of water.
  56. 56Why is the 'leak' from a neuron more complicated than the leak from a bucket?Have to take into account chemical gradient of ions which affects the propensity to diffuse/movement.
  57. 57At what value is there no ion 'leak'? How does this differ to the bucket analogy?1) When EL(amount of voltage of membrane)=-70mV
    2) In the bucket analogy leak is equal to zero when the volume of the bucket is zero.
  58. 58How does the movement of ions initially appear counterintuitive?Would expect K+ ions to remain inside the cell at a membrane voltage of -60mV however due to high ion concentration inside the cell and the random Brownian movement of ions the electrical attraction is counteracted/balance - electrochemical gradient.
  59. 59What is the reversal potential?The voltage required for zero current even if there is some conductivity.
  60. 60What is the movement of current out of a cell proportional to?V-EL
  61. 61What does G from the bucket equation become?Gm - conductance, which measures the porousness of the membrane to the flow of ion.
    Constant of proportionality for the leak current.
  62. 62What equation denotes the leak current?Gm(V-EL)
    ie conductance x driving force
  63. 63How is resistance of the membrane obtained? What is the resulting equation?1) Divide across Gm(V-EL) as resistance is equal to 1/G
    2) Rm=1/Gm
  64. 64What is the T constant for a cell?Tm = Cm/Gm
  65. 65What is the final equation to describe the rate of change of voltage of a neuron?What is the final equation to describe the rate of change of voltage of a neuron?
  66. 66What is Tm? How is I traditionally denoted?1) Time constant
    2) To match the in vitro experiment - eg Ie where I is an injected current from an electrode.
  67. 67What does the equation leave out? How is this solved? What effect does this have?1) Leaves out important nonlinear changes in current as V changes eg during an AP.
    2) Add a spike 'by hand'
    3) Changes the voltage to a reset value - mimics what happens in a neuron/hodgkin and huxley model.
  68. 68What happens if V ≥ VT?Spike is recorded and the voltage is set to a rest value of VR.
  69. 69What is the voltage equilibrium value?Depends on neuron, V is = EL + RmIe
    eg the RMP plus input voltage.
  70. 70—
  71. 71When is the integrate and fire model useful? What does it lack?1) Modelling the behaviour of large neuronal networks
    2) Exploring ideas about neuronal computation in a relatively straight-forward setting.
    Lacks lots of the important detail in the dynamics of neurons.
  72. 72How is the Tm(dV/dt) equation solved? What is a requirement for this solution to be correct? What happens if this is not the case?Ie must be constant - Can be solved numerically on a computer.How is the Tm(dV/dt) equation solved? What is a requirement for this solution to be correct? What happens if this is not the case?
  73. 73Given a constant input/current Ie - What is the value where V stops changing (equilibrium value)?Tm(dV/dt) = EL - V +RmIe
    When dV/dt = 0
    0=EL - V + RmIe    Given a constant input/current Ie - What is the value where V stops changing (equilibrium value)?
  74. 74What are the possible outcomes when a neuron voltage reaches towards its equilibrium value?If equilibrium voltage (Ve) reaches threshold voltage (Vt) the neuron will spike repeatedly.
  75. 75How could you work the relationship between firing rate and current? What would you expect to see in terms of firing rate due to different input voltages?1) Use the constant Ie to work out the f-I curve - ie the relationship between firing rate and input current
    2) The greater the input voltage (and therefore the current) the higher the firing rate although only once threshold has been reached.
  76. 76—
  77. 77What do the nonlinear dynamics of voltage spikes arise from? What is a crucial feature of theses channels?1) Voltage-gated channels (conductance varies as voltage varies)
    2) Selective therefore different types currents for each type of ion, each is non-constant.
    Complicated nonlinear function
  78. 78Which currents form the basis of the Hodgkin-Huxley equation? What does the equation describe?1) VG channel currents
    2) Equates the rate of change of V to a set of currents. The roughly linear below threshold voltage is set to the leak currents. The nonlinear behaviour of voltage/current near threshold is assigned the voltage gated current.    Which currents form the basis of the Hodgkin-Huxley equation? What does the equation describe?
  79. 79What type of model is the Hodgkin-HUxley model referred to as? Why?Conductance based model - currents depend on conductance and conductance depends on gated channels.
  80. 80What describes the conductances of VG channels?Gating equations.
  81. 81Describe the characteristics of the Hodgkin-Huxley model? (2)1) One voltage
    2) Pretends each neuron is a simple point
  82. 82What can more complicated models include?Spacial extent of neurons and model the neuron as being composed of compartments - each compartment consisting of a Hodgkin-Huxley equation.
  83. 83What describes how the compartments are linked together?Cable equation.
  84. 84What are the more general computational tools?MATLAB and Python
  85. 85What are the more specialised tools? What are they used for?1) GENESIS - Large complicated neurons
    2) NEURON - " "
    3) NEST - Large complicated networks of neurons.
  86. 86Which two computational tools have an interface?1) MATLAB
    2) Python
  87. 87What are the advantages of MATLAB?1) Very large user base
    2) Huge collection of libraries and packages
  88. 88What is the free community version of MATLAB? What is the difference?Octave - the two packages aren't completely compatible and Octave lacks many of the packages and libraries
  89. 89What type of paradigm does MATLAB use? What is the advantage of this?Matrix paradigm - easy and quick when you get used to it
  90. 90What is Python? What are it's advantages? Disadvantages?1) Programming language
    2) + Nicer to use than Matlab
    + is free and open
    + Many libraries
    3) - Fewer libraries than Matlab
  91. 91What is Brian?Python package specialised for writing neuron simulations
  92. 92What is NEURON used for?Large biophysical simulations of neurons
  93. 93What is GENESIS?Rival to NEURON
  94. 94What is NEST used for?Simulating very large networks of simple neurons
  95. 95What is XXP?Powerful tool for investigating dynamical systems of the sort found in neuroscience
  96. 96What is Open Source Brain?A set of tools for sharing neuronal simulations
  97. 97What is NeuroMLMarkup language for neuron models

Información

Descripción
Year 2 Techniques in Neuroscience Fichas sobre Computational Neuroscience , creado por Anna Hogarth el 16/04/2016.
Creado por
Anna Hogarth · abr 2016
Actualizado
28 dic 2016
Visitas
2
Etiquetas
cian o'donnel
Valoración

Más sobre Computational Neuroscience and Neuron Models

Escríbenos

¿Sobre qué es tu mensaje?

¿Qué te parece el nuevo diseño?

¿Te gusta?