Laboratory of Computing Biological Networks - group meeting

Information in the Retina and Experiments on Photon Counting

Kevin Sean Chen — Fri, 24 Jul 2015 16:58:17 +0000

There are mainly two parts in my presentation this time, results and analysis for the omitted stimulus response (OSR) experiment with spatial stimuli and some preliminary tests for photon counting in retina.

Last time, I talked about a hypothesized simple columnar organization that produces OSR in the retina. We may investigate the circuits by providing light stimuli with spatial patterns. Uniform, checkerboard, and random distributed pixels were used as the periodic stimulus. Fixing the number of flashes, period, and the average brightness, we see that OSR can only be generated under uniform, checkerboard or fixed random pixels (that are unchanged within a trial), but not in the anti-checkerboard or random pixel stimuli (that changes in every flashes). This implies that there may be specific spatial units to produce OSR. Further investigation is needed to identify the size and properties of these units.

In addition, I calculated spatial correlation and temporal information from the spike trains in previous experiments. For spatial correlation, we find that the correlated distance is longer under uniform stimuli than random flickers or checkerboard. Interestingly, those producing OSR seem to also have shorter correlation distance than those that simply response to each light flashes. For temporal information, my extrapolation results seem to provide less entropy rate comparing to the previous researches. I might find another method that fits our experiment better to calculate temporal information. Lastly, I’ve been reading articles focusing on the strong inhibitions in the retina. The inhibitory signal may not only play an important role in the anti-Hebbian model but also help encoding information through silencing neighbor neurons. Since we've been recently discussing the strong inhibition right after OSR, I'm also interested in how OSR affects the synergitic codes in retina.

For photon counting on the retina, I briefly went through some classic experiments showing the single photon sensitivity and detection of photon statistics in photoreceptor cells. Most researches focus on the gain control and regulation of threshold in the retina under dim light. In the future, we hope to calculate the correlation between spikes from ganglion cells, which really send information to the brain, to investigate whether photon information could still preserve after processing through the retinal circuits.

Tags:

retina

OSR

Information

Photon Counting

group meeting

Progress, future works on OSR, and some discussions on retina wave

Kevin Sean Chen — Sat, 06 Jun 2015 06:49:05 +0000

Some results from our experiments on OSR were shown in the slides. By holding the last state after terminating the periodic stimulus in an inverted manner (differ from the originally adapted brightness), OSR could be eliminated or shifted significantly in the latency. We’ll try to repeat the experiments a few more times, since such designed stimuli would tell us more about how ON and OFF pathways in the retina interact to produce OSR.

In the hypothetical circuit in the retina to perform predictive coding as observed in the experiments, I think gap junctions might play an important role to also modulate the activity in space. Furthermore, I show the raw data before passing the filter or detecting spikes. We can guess that a strong, synchronized inhibition is generated when OSR occurs. To sum up, I plan to perform some spatial and temporal analysis according to some references from a group also working on frog retina. Also, flashing bars or checkerboards to test the affects on OSR may also be a good idea to investigate the mechanisms for dynamical predictive coding and analyze the circuit in retina.

In the second part, I reported an article about the segregation and desynchronization of ON and OFF pathways during a certain stage in retina waves. The organization of ON and OFF paths may affect the connection in LGN and also some important orientation selectivity in V1 cortex. In result, ON bipolar sends lateral inhibition to OFF bipolar cells through inhibitory neurons like amacrine. Also, controlling the reuptake of glutamate by Mulller cells in the retina also affects the temporal precision for retina waves. Last but not least, most of theses cells are connected through gap junction, which also participates in strong inspatial synchrony. Since we're now able to recieve signals from the embryonic retina, we're optimistic to repeat some experiments in the refernce and try to stimulate it with different dynamical light patterns.

Tags:

retina

Omitted Stimulus Response

Predictive Coding

Retina Wave

group meeting

Firing rate model (Mean Field theory) for neural population

Ying-Jen Yang — Fri, 08 May 2015 08:06:15 +0000

Mean Field Theory for a neural population is the lowest order macroscopic description for a neural network. By defining the firing rate probability density of the network, a dynamical equations can be written to describe the system. In usual case, there are two importent element in the equation : the time scale and the gain function. To understand the microscopic correspondence, I introduce the central idea of the derivation.

First, one must know that there are two coding strategy in time for neural system : temporal coding and rate coding. According to Taillefumier and Magnasco's work, there is a transition between them depend on how "noisy" the input is comparing to the intrinsic noise which makes the system a random walker. In the rate coding region, one can use firing rate to describe a neuron or a neural network. One can use Linear-Nonlinear model to link the stimulus to the response (firing rate). In a network, a mean field theory can describe the neural population by the firing rate. By studying the stationary state, where all neuron fires in a Poisson process with same firing rate. The mean firing rate can be found by the gain function. For the dynamical equation, there are lots of derivation. One can either read the famous Wilson-Cowan model or read Chap.12~14 in Gerstner et al.'s book. In a word, the time constant is actually the natural time scale for a single neuron. And for the gain function, it is related to the degree distribution and also the firing rate- input current relation. ( In a all to all network, the gain function is the firing rate-input current relation.)

In practical concern, one treat the dynamical equation as a phenomenological model, there are several often-used gain function. Linear, threshold linear or the most physiological one -- sigmoid. Just to remind you that the first two is just a low activity approximation since neuron can't fire infinitely fast. There must be a cutoff in the high activity case. Last but not least, Jack D. Cowan established a statistical field theory for neural network which can also consider the fluctuatio and also the correlation between the cells or between populations. One can find the history in Chap.2 in his latest published book.

One can find my slide here.

Tags:

Progress report on EGTA and OSR measurement in retina

Kevin Sean Chen — Sat, 02 May 2015 07:01:46 +0000

During the last two months, we started adding EGTA, a calcium chelator in the buffer and observed change in OSR. Surprisingly, period doubling occurred at some specific concentration, and didn’t abolish OSR. We think that EGTA affects the time scale in the whole retina activity.

Following up, we designed light stimuli with two periods and also produced activity similar to period doubling, then measured the latency of OSR. In fact, the latency seems to be highly dependent to the last pulse. We can control the latency through the brightness of the last pulse. This may be a useful method for our future work, to do real-time feedbacks on retina for latency control.

Tags:

Literature Reivew : "Broad-range" Stochastic Resonance

Ying-Jen Yang — Mon, 30 Mar 2015 06:07:52 +0000

Stochastic Resonance(SR) is a phenomena that a excitable system is capable of detecting weak signal best with some optimal noise strength. In usually case, the optimal noise strength is only one value (one maximum in the S/N versus noise strength plot). Since our environment is always changing, a crucial problem is that is the noise strength always in this optimal level ? Namely, for different environment, thus different noise strength, does our system automatically tune the excitability such that it is optimal ? This means that we will have a flat plateau in the S/N versus noise strength plot(say, SR curve)! The motivation now is clear. We have developed a adaptive FitzHugh-Nagumo model that can tune its excitability. This model should have non-trivial behavior on SR and even Coherence Resonance (CR)! Therefore, let us see if there is some clues in the past research. In other words, what kind of system have a nontrivial SR curve.

A paper published at Nature in 1995 considered the signal passes many parallel FHN systems and sum up there responses as the output. There main finding is that when the number of parallel FHNs become large, there is a plateau in the SR curve! Also, these noise added will not affect the ability to detect supra-threshold signal. Another similar system published at Plos one in 2011 considered several independent noise sources producing Poisson firing by presynaptic neurons affect the postsynaptic neuron through TM dynamical synapse (with both synaptic depression and facilitation). And also added a weak signal to the postsynaptic neuron. The results showed that with these kind of “synaptic filter” tuning the noise effect. There will be two peaks in the SR curve and the place where peak happens can be tuned by changing the time scale of synaptic plasticity. These two systems with nontrivial SR curves are both “prefilter”-like. I also found two PRE papers which is more alike with our adaptive FHN system

One published in 2008 was Volman's previous work. They consider a random neural network with N neuron and the synaptic dynamic is the same as the bursting paper we are familiar with. There main founding is that asynchronous release, that is a dynamic-dependent noise source, will make the SR curve changes with N. And turns out to have optimal N and optimal asynchronous release strength g for weak signal detection. The other paper in 2014 was considering a different excitable system with positive feedback and negative feedback. They tried to work out the effect of this +/- feedback on “CR curve”, how coherence of the noise-driven activity versus the noise level. I think after analyzing the SR behavior of our adaptive FHN model, we can compare our results with these two models.

One can find the html I presented here : https://networks.tir.tw/~yjyang/present/GM1tuningSR.html

The four papers I mentioned in this talk are :

Tags:

Stochastic Resonance

Short-term Synaptic Plasticity

Adaptive Excitability

group meeting

Spatiotemporal Memory Is an Intrinsic Property of Networks of Dissociated Cortical Neurons

Kevin Sean Chen — Fri, 20 Mar 2015 18:20:10 +0000

Following Tina's presentation last time, this article was published by the same group this year and they moved on to investigate spatiotemporal memory properties in the cultured neural network. I think some ideas and results in the article may be informative for the work on retina and the future works in cultured neurons in our group.

The cortical neurons were cultured on MEA, transfected by ChR2 (an optogentetic light-sensitive ion channel), and stimulated by laser patterns generated through spatial light modulator. Four jittered light pulses and forty different "music" (4 "notes" presented with different spatial patterns) stimuli were design to verify the ability for neural network to memories and classify these inputs, and the classification accuracy for output spikes from the MEA recordings were compared with the liquid-state machine.

In result, they conclude that cultured neural networks are able to accumulate information and eventually classify different types of light stimuli, utilizing spatial temporal information. Such ability to maintain information to classify input is an intrinsic property. In addition, they also indicate how bursts might wash out information for classification and also how STSP may be important for spatiotemporal memory in the network.

Tags:

Spatiotemporal Memory

Short-term Synaptic Plasticity

Optogenetic

Liquid-State Machine

group meeting

Experimental Studies of Anticipative Dynamics in Neuronal Networks

José Wu — Fri, 06 Mar 2015 07:39:48 +0000

Abstract

Understanding how neural systems integrate sensory signals to perceive time is a fundamental problem. However, the basic neuronal mechanism of temporal perception is still far from clear. Currently, there is a debate on whether a specialized clock is needed for time perception (TP). Our view is that TP can be a general property of neural networks endowed with short term synaptic plasticity (STSP) and enough recurrent connections. Recent simulation studies of anticipative dynamics (one form of TP) support this later mechanism. The goals of our proposed experiments are: i) to demonstrate that STSP is crucial in the anticipative dynamics in the retina of frogs and ii) to induce anticipative dynamics in a cortical neuronal culture of rats by using a recently invented photo-sensitive technique based on the organic semi-conducting polymers, P3HT. Since there are no special clock circuits in the retina and the cortical culture, results our proposed works will help to clear some of the controversies in the perception of time.

Reference

[1] A. T. Winfree, The Geometry of Biological Time, Springer, (2001) 19.

[2] R.B. Ivry, J.E. Schlerf, Dedicated and intrinsic models of time perception, Trends Cogn Sci, 12 (2008) 273-280.

[3] G.Schwartz,R.Harris,D.Shrom,M.J.Berry,II,Detection and prediction of periodic patterns by the retina, Nat. Neurosci., 10 (2007) 552-554.

[4] R.B. Ivry, R.M. Spencer, H.N. Zelaznik, J. Diedrichsen, The cerebellum and event timing, Ann N Y Acad Sci, 978 (2002) 302-317.

[5] V. Jacob1, L. Petreanu, N. Wright1, K. Svoboda, K. Fox, Regular Spiking and Intrinsic Bursting Pyramidal Cells Show Orthogonal Forms of Experience-Dependent Plasticity in Layer V of Barrel Cortex, Neuron,

[6] P.M. Lau, G.Q. Bi, Synaptic mechanisms of persistent reverberatory activity in neuronal networks, Proc. Natl. Acad. Sci. USA, 102 (2005) 10333-10338

[7] V. Gautam, D. Rand, Y. Hanein, K.S. Narayan, A Polymer Optoelectronic Interface Provides Visual Cues to a Blind Retina, Adv. Material (2014) 26:1751-1756

[8] L. C. Jia, M. Sano, Pin-Yin Lai and C. K. Chan, ”Connectivity and Syhchronous Firing in Cortical Neuronal Networks ,Phys. Rev. Lett. (2004) 93, 088101

[9] U.R. Karmarkar, D.V. Buonomano Timing in the absence of clocks: encoding time in neural network states. Neuron (2007) 53, 427–438

Tags:

Anticipative dynamics

Optimal Heterogeneity for coding in spiking neural network

Ying-Jen Yang — Mon, 26 Jan 2015 13:49:17 +0000

Today I shared a PRL paper considering the Heterogeneity in integrate-and-fire neurons' threshold.

http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.108.228102

They consider a all-to-all excitatory integrate-and-fire network and varied the deviation of their thresholds. What they found is that a proper level of heterogeneity will introduce low-threshold neurons that enhance the activity and sensitivity of the all-to-all network. This kind of enhancement leads to a proper level for the efficiency of rate coding and spike-time coding.

Also, in their more recent efforts, they found that the heterogeneity effect in inhibitory population is different from the one in excitatory population in their previous consideration.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4162374/

As CKC have comment, this kind of proper heterogeneity level may be quite specific to the all-to-all network case. For other network topology, this might not be true. Also, the threshold is actually a time variable associated with time in the network with synapitc plasticity.

Tags:

Heterogeneity

rate coding

spike-time coding

spiking neural network

group meeting

Simulation of spontaneous bursting activities

justinhuang — Mon, 12 Jan 2015 10:35:28 +0000

The spontaneous bursting activities can be widely observed in many different experiments about the development of neuronal networks. One typical example is the periodical spontaneous bursting activity during development of the in-vitro cortical neuronal network from embryonic rat's brains observed through the MEA device. Currently, I am working on the simulation of this type of neuronal activities and, based on our simulations, try to understand the generation mechanism of what we have observed from experiments. The details and reports about my works can be found in my <a href="https://networks.tir.tw/~justin/"> personal homepage< /a> in this website.

Tags:

MEA

spontaneous bursting activities

Volman's model

group meeting

Single Action Potentials and Subthreshold Electrical Events Imaged in Neurons with a Fluorescent Protein Voltage Probe

José Wu — Mon, 12 Jan 2015 09:55:47 +0000

Paper site

A genetically encoded sensor of membrane potential, FlaSh, was first introduced by Siegel and Isacoff (1997) as a fusion between the Shaker potassium channel and wild-type green fluorescent protein from Aequorea victoria (aqGFP).

Subsequent ion channel-based voltage sensors were designed to include a single fluorescent protein or FPs that form Förster Resonance Energy Transfer pairs (FRET).

Later sensors based on the voltage-sensing domain of Ciona intestinalis voltage-sensitive phosphatase (CiVSP) produced robust signals in mammalian cells.

These group have combined many Ciona intestinalis voltage sensor (CiVS) with different FPs to produce FP voltage sensors with improved properties.

However, to date this approach had not yielded probes with the necessary combination of signal size and speed that would make it possible to image individual voltage signals in neurons:
1. Action potentials
2. Subthreshold potentials

This paper report the development of an FP voltage sensor, named ArcLight, which is based on a fusion of the CiVS and the fluorescent protein super ecliptic pHluorin that carries an A227D mutation.

Tags:

CiVP

ArcLight

Voltage sensitive dye

group meeting