Computational Neuroscience

Question 1

Brain model

Answer 1

Emulate: to develop algorithms
Heal: find treatments
Understand: how brain works
Brain has ~86 billion neurons
Neurons connected to ~10,000 other neurons
Neuroscience has lacked behind due to complexity and lack of data but catching up
Biologically good computational heavy

Question 2

Scales

Answer 2

Macroscale: a whole region
Mesoscale: interaction between a few regions
Micrpscale: 1 neuron

Question 3

Allen Brain Atlas

Answer 3

Need normal human brains: natural death
MRI of brain
Diffusing tensors to see the connections
Slice of brains : 3D model to get anatomy
repeat and slice + fragment.
Darker stains are dense in neurons
Microarrays on thousands of samples: human genome
Have modeled complete brains

Question 4

Neurons

Answer 4

Specialized to generate and conduct electric signal
Dendrites to receive input
soma: cell body
Axion to carry output. Covered of myelin sheet: best conductance, AP don’t have to regenerate (saltatory conduction).
Resting potential: potential inside neuron compared to extracellular medium ~70mV
High concentration of patassium (K+) inside neuron
High concentration of sodium (Na+) outside neuron
Ion channels open and close (also of Cl- and Ca2+) by release of neurotransmitters
Neurons respond to different orientations (tuning curves)

Question 5

Squid neuron

Answer 5

Squid have giant axon
Useful to study action potentials, initiation, propagation -> model ion channels and synapses
Start of neuroscience field

Question 6

Action potential

Answer 6

spike, rapid rize and fall. Depolarization followed by Repolarization/Hyperpolarization. Occurs if reaches firing threshold.
absolute refractory period: channels closed, impossible to have another AP
relative refractory period: difficult, need bigger stimulus, but possible
AP regenerate at each synapse

Question 7

Inhibitory/Excitatory potentials

Answer 7

IPSP: Potassium(K+), Chloride channels(Cl-) -> repolarization
EPSP: Sodium (Na+) -> depolarization
PSP: graded potential, sum
Passive conductance of PSP from dendrites to soma to axon
sum EPSP - sum IPSP > threshold to have AP

Question 8

Hodgkin & Huxley model

Answer 8

Conductance based
4 differential equations to describe AP
Sodium, potassium, small leak (Cl-)
Calculate currents, conductance and voltage through cell
Strength stimulation vary the shape and amplitude
Fast spiking neuron, regular spiking neuron (compare fast and regular by time), bursting neuron (multiple AP then none)

Question 9

Integrate and fire neuron

Answer 9

large number of neuron possible
subtreshold leaky integrator
firing treshold
reset

Question 10

Compartmental model

Answer 10

Most detailed
Take shape, size, density properties in different parts, …
detailed simulation of a few neurons

Question 11

fMRI

Answer 11

correlation and high spatial
blood oxygenation level dependent signal
signal intensity depend on oxygen level: more active -> more oxygen consumed -> more blood flow
same technique as MRI
Change in activity during task
Mapping activity: activity correlation through time, how much coactivate regions in matrix (correlation strength

Question 12

Population coding

Answer 12

Neural spike noisy
data from many trials to see correlation
population average: brain don’t have many trials
Overage over time in repeated experiments as subtitute. Divide populations in to subpopulations of same type, similar response, etc -> describe activity of mean neural population rather than individual neuron spiking
Neurons are organized in populations of same properties

Question 13

Classification neuroimaging

Answer 13

Temporal to spatial resolution
Capacity to establish correlation or causal relation
Correlation: compare activity with and without light, not sure if all activity necessary to perceive light
Modulated: put light directly on visual field. Sure the consequence is because of it.

Correlation with high spatial resolution: MRI and fMRI
Correlation with high temporal resolution: EEG, MEG
high spatial have low temporal and the opposite. Measure activity indirectly >< directly

Interferential (modulate brain) : TES, microstimulation, optogenetics (only animals)

Question 14

MRI

Answer 14

correlation and high spatial
Magnetic Resonance Imaging
big magnet that adds perturbation with coils. See how tissues are affected by changes in magnetic field -> different tissue properties
Give anatomical images or videos
Different parameters can visiualize different thins: vasculature, ion accumulation, etc
Mapping anatomical connectivity: undirected, proportion of fiber tracts in matrix

Question 15

Network neuroscience

Answer 15

analyze brain networks
Brain network between region.
Adjency matrix indicate strength of connection: how many axons interconnect.
Directed graph: direction of information flow. anatomical too small to find

Question 16

Mapping information flow

Answer 16

energy one node, see how it spreads in network with EEG/MEG

Question 17

Brain network structure

Answer 17

balance between functional segregation and functional integration
There are hubs and communities
Functional segregation: neurons processing similar things are clustered together
Functional integration: efficient communication between modules
Rich club organization: hubs connect the population but hubs also connect to the other hubs (not the case for PPI). Many short paths/ Makes it more robust and flexible in case one would deteriorate ! Worst degradation is on Rich club because impacts everything.

Question 18

EEG/ MEG

Answer 18

Correlation and high temporal
Electro/Magneto encephalography
Pick up current of neuronal activity
Fast/slow frequency indicate activity
Mapping information flow

Question 19

Transcranial stimulation

Answer 19

TES/TMS
Transcranial electric, negative and positive pole on either side. Weak pulse: only affect excitability, not enough to fire
Transcranial magnetic: more precise and stronger, AP generated.
TMS to stimulate, EEG to monitor activity
Influence modulation
Mapping information flow, directed information spread by activating a node

Question 20

Microstimulation

Answer 20

electric currents via electrode implanted in the brain

deep brain stimulation, can apply stimulation with a different pattern: parkinson treatment

Question 21

Optogenetics

Answer 21

on animals
manipulate AP with light sensitivity. More precise than electrical stimulation
inject a virus that will be activated by light due to a sensitive protein

Question 22

Neuroprosthetics

Answer 22

Device that can be controlled by input of nearby muscle or CNS
Some read sensory input

Question 23

Perceptron

Answer 23

mathematical representation of neuron input/output
w= vectors of weights (should be trained)
b = bias (firing threshold)
n = net input (sum Weights + bias)
f = activation function (generation AP)
a = activation value, neuron output ( AP)

Question 24

Training neuron network

Answer 24

weights updated by back propagation; error estimation, change in parameters

Question 25

Sensory discrepancy

Answer 25

Difference prediction and output

Question 26

Internal forward model

Answer 26

simulate response of system to input signal u(t). -> predict consequences of motor movement

Question 27

Efference/Reafference

Answer 27

Efference: prediction of movement. need input signal if not then prediction difficult.
Reafference: sensory prediction. The prediction is cancelled when performing task
If not direct force: applied on object, then not cancelled by sensory imput
Thus if unexpected -> amplified
There is a sensory processing sensitivity that loops sensory information
Perception of force reduced by 0.5 when done ourselves. Self generated always perceived as weaker.

Question 28

Systems Biology >< Computational Neuroscience

Answer 28

SB: data rich, infer unknowns
CN: data poor, simulations