# ankursinha.in/blog

neuroscience/fedora/musings

Wed 29 July 2015

# CNS 2015 - Day 0

Posted by ankur in Research (1207 words, approximately a 5 minute read)

The notes have not been proofread. Please do your research before you pick anything from this post. It is meant to be a rough sketch of everything that I heard and noted at the conference. Since quite a bit of this is new to me, it is bound to be inaccurate, unspecific, and possibly even incorrectly quoted.

## Conference details

The conference was between July 18 and July 23, 2015 and was at the University of Economics in Prague. More information on the conference can be found at the website here.

## Arrival in Prague

We arrived by a slightly late flight at the Prague airport. Even before we'd left the airport, my supervisor, Volker, introduced us to Eugene Izhikevich at the baggage carousel. I was obviously star struck and mentally noted how well the trip had begun. We took a bus and a train to get to our hotels which were all around the conference venue in Prague 3.

## Day 0 - tutorials

I initially started at the Brian tutorial hoping to learn more about the simulator. The tutorial was more about using the simulator, and since I'm not using the simulator at the moment, I moved to the tutorial on Calcium Dynamics. The tutorial was most informative, and of course, a rather large part of the information was completely new to me. Here are the notes I scribbled in my not so scientific terms:

1. Calcium is critical for plasticity.
2. Kinase takes a phosphate group and puts it on to another molecule.
3. CaMKII is required for LTP but not required nor activated in LTD.
4. Inhibition of CaMKII blocks LTP and memory.
5. Phosphatases play a role in LTD, for example PP2B (Calcineurin) is also activated by Ca.
6. Both LTD and LTP require elevation in Ca++, so how does the Ca++ decide whether to bind to CaMKII (LTP) or PP2B(LTD)?
1. Calcium calmodulin first binds to PP2B.
2. With higher concentrations of Calcium calmodulin, the excess Ca binds to CaMKII.
3. The CaMKII activity overcomes PP2B activity.
7. The dynamics of Ca and CaMKII do not explain everything.
1. In neocortical Calcium neurons, for example, the magnitude of peak Calcium does not always predict the direction of plasticity.
2. CITE: Nevian and Sakmann - J. neuroscience 2006.
8. Other protein kinases:
1. Protein kinase C
2. Protein kinase A and so on.
9. E-LTP - early LTP
10. L-LTP - late LTP
1. These names aren't the best since people sometimes confuse themselves by thinking that the L means "late induction", which is not the case.
11. Signal pathways and LTP:
1. Neurotransmitter -> receptor -> second messenger -> kinase/phosphatase -> nucleus.
12. It is yet to be shown that STDP is protein synthesis dependent.
13. General method of studying effects of molecules seems to be the introduction of inhibitors and analysis of their effects.
14. MAPK activation is reqired for L-LTP and fear learning.
15. There are changes in the dendritic spine.
16. Synaptic plasticity, excitability, gene regulation and memory are all controlled by intracellular signalling pathways.
17. G protein coupled (metabotrophic) receptors are involved in LTP.
1. Direct action - directly gates channel
2. Indirect action - via downstream messengers.
3. For direct action, the channels that are closest are acted upon.
4. For indirect action, G protein binds to an enzyme -> enzyme produces second messenger -> therefore, wide spatial extent due to diffusible second messenger.
18. G proteins can be heterotrimetric:
1. They are made up of three subunits: alpha, beta and gamma.
2. The beta and gamma subunits bind to the alpha subunit and prevent it from binding.
3. GDP - inactive.
4. GTP - active.
19. Enzymes activated by G proteins:
2. Phospholipase C (Gq)
20. Stoichiometric reactions - the equations provide information on the number of molecules required
21. The transition from closed to open channels are a first order chemical reaction:
\begin{align*} substrate &\underset{k_b}{\overset{k_b}{\rightleftharpoons}} product \\ p/s &= k_f/k_b \\ d[prod]/dt &= k_f[substrate] - k_b[product] \\ [substrate] &= [initial\_concentration] - [product] \end{align*}
1. Second order reactions:
\begin{align*} substrate1 + substrate2 &\underset{k_b}{\overset{k_b}{\rightleftharpoons}} product \\ d[prod]/dt &= k_f[substrate1][substrate2] - k_b[product] \\ [substrate1] &= [initial\_concentration\_substrate1] - [product] \\ [substrate2] &= [initial\_concentration\_substrate2] - [product] \end{align*}
1. Third order reactions:
\begin{align*} substrate1 + 2~substrate2 &\underset{k_b}{\overset{k_b}{\rightleftharpoons}} product \\ d[prod]/dt &= k_f[substrate1]{[substrate2]}^2 - k_b[product] \\ [substrate1] &= [initial\_concentration\_substrate1] - [product] \\ [substrate2] &= [initial\_concentration\_substrate2] - 2~[product] \end{align*}
1. Enzymatic reactions:
\begin{align*} Enzyme + substrate &\underset{k_b}{\overset{k_b}{\rightleftharpoons}} ES \overset{k_{cat}}{\rightarrow} Enzyme + product \\ d[ES]/dt &= k_f[Enzyme][substrate] - (k_{cat} - k_b).[ES] \\ dprod/dt &= k_{cat}.[ES] \end{align*}
1. Backward rate constant is very near zero.
2. Michalis-Menton Dynamics(?).
3. ES rapdly reaches equilibrium.
4. Substrate is in excess (enzyme quantity is rate limiting).
5. At equilibrium:
\begin{align*} d[ES]/dt &= 0 \\ k_f[Enzyme][Substrate] - (k_b + k_{cat}).[ES] &= 0 \\ [Enzyme] &= [Enzyme\_total] - [ES] \\ [ES] &= ([Enzyme\_total].[Substrate])/((k_b + k_{cat})/kf + [Substrate]) \\ d[product]/dt &= (k_{cat}.[Enzyme\_total].[Substrate])/([Substrate] + k_m) \\ k_m &= (k_{cat} + k_b)/k_f \end{align*}
1. $$k_f$$ is affinity - no need to know $$k_b$$ and $$k_f$$.
2. Note that at max:
\begin{align*} d[product]/dt &= k_{cat}[Enzyme\_total] \\ [Substrate]/([Substrate] + k_m) &= 1~as~[substrate] -> \infty \end{align*}
1. General rule:
\begin{align*} d[product]/dt = \sum\limits{all\_rxns} (k_f \prod{[Substrate]} - k_b \prod{[product]}) \\ \end{align*}
1. Software:
1. XPPAUT - general purpose ODE solver.
1. To smooth out sudden spikes, you filter it out through an RC circuit to make it smoother.
2. Just write ODEs normally.
3. "aux" key word if you want to see the value of variables not in differential equations.
4. Saves data to text files.
5. Good tool for single neuron analysis, like AdEx analysis - shows values of diff equations and so on.
6. Can also do nullclines.
7. Has many ODE solvers to choose from.
8. Also does phase plane analysis.
2. Genesis/Chemesis/Kinetikit/Moose #. Moose has a python interface #. use "showfield" to check values #. Basically, you make different molecule variables and so on and hook them up together at the end.
2. Smooth endoplasmic reticulum are Calcium stores
3. Sources of Calcium:
1. Calcium currents (L,N,P,Q,R,T)
2. Internal stores
3. Calcium sinks
4. Source or sink:
1. Buffers
2. diffusion
1. Working with experimentalists is a must for accurate constants while modelling Calcium concentrations.
2. Calcium current:
\begin{equation*} I_{Ca} = g_{max}m^ph^q (V_m - E_{Ca}) \end{equation*}
1. but Calcium currents don't reverse and it is not well understood why, so we use the Goldman-Hodgkin-Katz equation.
2. It applies when inside and outside concentrations are asymmetrical.
3. Influx due to Calcium:
\begin{align*} \phi &= (-I_{Ca}(Vm))/(charge.F) \\ F &= Faraday's constant \end{align*}
1. Flux = moles per unit time.
2. $$E_{Ca} > E_{Na}$$, so never have Ca flowing in. We must use indirect methods - 10% of NMDA conductance and so on.
3. Ion channels have multiple states one of which is "open and conducting".
4. Hodgkin-Huxley model says the different subunits are independent:
\begin{equation*} n^4 = n_1 \times n_2 \times n_3 \times n_4. \end{equation*}
1. So, it works for K: single channel data is sufficient for multi state model.
2. But, this doesn't work for Na channels, which is when you need to include different states of channels.
3. Pick the level of modelling that you need.
4. CITE: Li and Rinzi - equations to model Ca channel dynamics.
5. CITE: Scheuss et al 2002
6. Note: Calbindin and Parvalbumin are used to classify neurons.
7. Spines are generally modelled in a single dimension.
8. Genesis and Moose have simplified Calcium implementations.
9. TIP: always take a peek at the sources of implementations to make sure what units are used.
10. Volume of ER is about 18.5% of full cell.
11. Spatial stochastic simulators
1. Smoldyn, MCell, CDS
2. Each molecule is a point particle
3. Computations increase with the number of molecules
4. MCell:
1. Geometry from volumetric data using Blender
5. CDS:
1. Particle based simulators with event driven algorithm.
2. Investigate crowding and so on since particles have a volume.
3. CaMKII diffusion depends on morphology.
6. Smoldyn
1. Relatively easy to use and learn.
12. Non spatial stochastic simulators
1. Each simulator has different goals
2. NeuroRD - mesoscopic.
1. Different morphology file and reaction file.
2. Only first and second order at this time.
3. A separate initial condition file.
4. Stimulation file.
5. Output specification -> list of molecules.
6. Top level modal file that tells what files do what.
7. Java
3. VCell