Information Representation in Neurons

[jpaul]

Hello everybody,
So I've been studying how the brain represents and encodes information. There is ample evidence/info showing that neurons adjust their firing rates and strengths of their synapses in order to encode information and form accessible neural pathways. However I am having trouble finding literature on the role of molecules like neurotransmitters in information representation. I hope my problem is clear enough, thanks for reading!

 

[jedishrfu]

Welcome to PF!

For starters:

http://en.m.wikipedia.org/wiki/Neurotransmitter

 

[DiracPool]

Hello everybody,
So I've been studying how the brain represents and encodes information. There is ample evidence/info showing that neurons adjust their firing rates and strengths of their synapses in order to encode information and form accessible neural pathways. However I am having trouble finding literature on the role of molecules like neurotransmitters in information representation. I hope my problem is clear enough, thanks for reading!

If you want to know how the brain encodes information, you don't want to look at it at the level of neurotransmitters. That is a microscopic analysis. The brain encodes information at the mesoscopic and macroscopic levels. Also, don't be fooled by a high firing rate of individual neurons, these are test models that don't typically occur in vivo conditions. Although you can artificially drive an individual neuron to up to a 1000 pulses per second, the in vivo neuron typically doesn't fire at more than, say, half a dozen.

It's the collective action of hundreds of millions of neurons all firing at about 6 hertz that make the brain work. We recognize information representation as patterns of cell assembly attractors that are formed and learned via neurohormones and transmitters, but that's only for storage, they don't play any specific role in information representation beyond that, other than a general action in the firing of neurons.

 

[Torbjorn_L]

So I've been studying how the brain represents and encodes information.

My own interest has been in how the brain works and how likely it is that such functionality evolves (astrobiology slant). And as suggested in the thread, that has taken me to meso- and macroscales. (Neurotransmitters are more useful to grok phylogenies there.)

I am also wary of the use of "information", as it is a set of loose constraints on a system, and rarely contribute useful ... well, information ... on it. So, with a layman selected unfamiliarity with the territory I have found these items of interest to me:

1. A robust way of long range information transmission in dynamical neural networks is to place them near criticality, with a branching likelihood of 1. I.e. if they branch too eagerly your network is swamped, if they branch too rarely a signal dies. This paper has studied in vitro neural tissue to confirm that it is possible:

[ http://www.jneurosci.org/content/23/35/11167.full.pdf html ]

- Now the caution of DiracPool applies, if this is an actual mode used in vivo.

- Another caution is that they rely on power laws over a mere 1-2 orders of magnitude, where you can always find something close to it and imply "fractal scaling". Eminent statistician Cosma Shalizi has criticized this repeatedly and developed bona fide statistical tests for power law behavior. Not surprisingly perhaps, he found that half of the 'power law' results out there are as well or better fitted with exponentials. This paper didn't do any of those tests IIRC.

That said, I also found the initial whole-brain imaging results of activity too chaotic to be modeled with self-organized "avalanches":



That is one brain on crazy!

However, later repeats feed the brains with organized information (the simulated background stripes on top left, cues to the fish to move forward or backwards to "keep in place"), and then you get organized behavior:



[ http://www.wired.com/2014/07/neuron-zebrafish-movie/ ; a lot more details here]

So maybe self-criticality is one way that organizes parts of brains, to complement the more usual "channel" behavior of the brain stem. (A behavior that can be easily identified in the 2nd movie, by the way!) And here "information" puts a useful constraint for once.

2. On larger scales, on the way to the symbolic processing of combinatorial languages such as ours, it seems the cortex of vertebrates (and so likely the homologous mushroom bodies of arthropods) self-organizes symbol learning. Presumably evolving robust learning has enforced evolution of a specific structure.

That is a way to get around the generic problem with neural network learning, over-training. (I.e. that the network learns too many specific quirks of the training material, so can't recognize it in nature. For example, if only fed faces oriented normally during training, an upside down face photo would be classified as "without face".)

"In this article from the Proceedings of the National Academy, Rougier et al. demonstrate how a specific network architecture - modeled loosely on what is known about dopaminergic projections from the ventral tegmental area and the basal ganglia to prefrontal cortex - can capture both generalization and symbol-like processing, simply by incorporating biologically-plausible simulations of neural computation. ...

In particular, it had developed abstract representations of feature dimensions, such that each unit in the PFC seemed to code for an entire set of stimulus dimensions, such as "shape," or "color." This is the first time (to my knowledge) that such abstract, symbol-like representations have been observed to self-organize within a neural network.

Furthermore, this network also showed powerful generalization ability. If the network was provided with novel stimuli after training - i.e., stimuli that had particular conjunctions of features that had not been part of the training set - it could nonetheless deal with them correctly."

[ http://develintel.blogspot.se/2006/10/generalization-and-symbolic-processing.html ]

Intriguingly the symbol-like behavior stems from a self-organized map of the active memory storage nodes, the "shape" and "color" dimensions mapping as PFC units within the network. That is (handwavingly) reminiscent of this year's medicine Nobel Prize find of "place" and "grid" neuron assemblies of the part of the hippocampus used in mammals to find their way around.

FWIW, I have never found much use for "information" constraints re symbol/map handling. They seem to be their own thing(s), quite different from (say) a template of Shannon information channels that can transmit such symbols. (But again, layman here.)

 

[DiracPool]

That is (handwavingly) reminiscent of this year's medicine Nobel Prize find of "place" and "grid" neuron assemblies of the part of the hippocampus used in mammals to find their way around.

Yeah, in my opinion this was absurd to give these people a Nobel prize for this. I think there must have been pressure on the committee to award something for all the progress that we've achieved in the field but this wasn't it. This GPS of the brain stuff is laughable. And embarrassing in my opinion. There's a long history of literature on the hippocampus and entorhinal cortex and it's relation to sense of direction in mammals. Lynn Nadel did a lot of the original pioneering work on that structure and this goes way back to the early 80's. I don't know why they left him out. There's this irresistible urge to compartmentalize brain functions to specific brain organs like the hippocampus, which isn't how the brain operates. I just cringe when I see statements like the GPS of the brain is located in the hippocampus. It's not that simple. As an historical note, you know they also gave the Nobel prize to Egas Moniz for the prefrontal leucotomy/lobotomy. That was a mistake. This gift to O'keefe and company isn't so egregiously missplaced as that, but it smacks of the same ignorance of who's contributions to neuroscientific research have been more relevant than others, again, IMHO.

Torbjorn is on the right track here. "On larger scales, on the way to the symbolic processing of combinatorial languages such as ours, it seems the cortex of vertebrates (and so likely the homologous mushroom bodies of arthropods) self-organizes symbol learning."

Brain attractors are self-organized and this leads to the self-organization of symbol manipulation, although it's too complicated to go into here. Mushroom bodies in arthropods are kind of the equivalent of the cortex in mammals and I was prepared to write a major review article on the evolution of modal action patterns and central pattern generators in arthropods and link it to the evolution of mammalian cortical dynamics, but regrettably that fell through. Thanks for reminding me Torbjorn.

"That said, I also found the initial whole-brain imaging results of activity too chaotic to be modeled with self-organized "avalanches"

Well, Prigogine's concept of self-organized criticality is of central relevance in mammalian chaotic dynamics. It's the avalanches and the critical masses that drive the alpha-theta rhythms globally for the large attractor states that govern cognition. I said in an earlier post that the typical cortical neuron fires at about 6 hertz, but it might even be less than that, maybe 2 to 3 hertz. In vivo, we see that the brain tends to keep neurons at suprathreashold state, letting this one and then that go over the edge to drive some sort of larger chaotic state. Again, it's the collective activity of hundreds of millions of these going on simultanously that drive that chaotic itinerancy. The self-organized criticality or "avalanche" states you refer to really work their magic by creating an order parameter of sequenced frames of either behavioral routines or symbol-manipulative cognitive schemes. That's how it works, it's really reminiscent of a Carnot cycle where the patterned activity of the suprathreshold neurons create a global attractor that act's back down on the subcortical structures to drive behavior.

http://www.ncbi.nlm.nih.gov/pubmed/23333569

 

[madness]

Yeah, in my opinion this was absurd to give these people a Nobel prize for this. I think there must have been pressure on the committee to award something for all the progress that we've achieved in the field but this wasn't it. This GPS of the brain stuff is laughable. And embarrassing in my opinion. There's a long history of literature on the hippocampus and entorhinal cortex and it's relation to sense of direction in mammals. Lynn Nadel did a lot of the original pioneering work on that structure and this goes way back to the early 80's. I don't know why they left him out. There's this irresistible urge to compartmentalize brain functions to specific brain organs like the hippocampus, which isn't how the brain operates. I just cringe when I see statements like the GPS of the brain is located in the hippocampus. It's not that simple. As an historical note, you know they also gave the Nobel prize to Egas Moniz for the prefrontal leucotomy/lobotomy. That was a mistake. This gift to O'keefe and company isn't so egregiously missplaced as that, but it smacks of the same ignorance of who's contributions to neuroscientific research have been more relevant than others, again, IMHO.

Sort of agree, in that we still don't really know that the hippocampus is really performing GPS like navigational functions, and there is still massive debate over it's function - the research community is divided between those who study its role in navigation and those who study its role in memory.

On the other hand, it's undeniable that O'Keefe and later the Mosers discovered novel cell types with interesting properties. These cell types are deep within the brain, far from sensory input, but their activity is highly dependent on external and behavioural factors, making them of clear interest for understanding higher level nonsensory processing. Secondly, they have spawned a huge research community with whole journals and conferences dedicated to it.

Hence, while I think the understanding of brain function that O'Keefe and the Mosers have achieved was vastly exaggerated by the Nobel prize committee, their contribution to neuroscience as a field was not at all exaggerated and was worthy of recognition.

 

[Torbjorn_L]

This GPS of the brain stuff is laughable. And embarrassing in my opinion. There's a long history of literature on the hippocampus and entorhinal cortex and it's relation to sense of direction in mammals. Lynn Nadel did a lot of the original pioneering work on that structure and this goes way back to the early 80's. I don't know why they left him out. There's this irresistible urge to compartmentalize brain functions to specific brain organs like the hippocampus, which isn't how the brain operates.

Thanks for the encouraging response! The rest was interesting, but I hope to see some tests of the notions eventually. I guess I have to reach for a review to get some notions of where the field is. I have been more interested in the earlier phases of biology such as emergence of life, that hominids also interest me is unnecessary stress.

I have to agree that "GPS of the brain" was embarrassing, but perhaps for other reasons than internal to the field that I know little about so far (especially its history). GPS are global and absolute coordinates, the "grid" assemblies are local and relative I think. Someone jumped on the erroneous but more exciting terminology.

The problem with selecting max 3 people gets worse every year, and it would have been surprising if this prize didn't labor under it. I read commentaries on that the Mosers have long been doing exceptional work in and of itself, and they were 2...

Agree on compartmentalization, which must be tempting, but - now we have zebra-fish! :w

 

[Pythagorean]

There are several basic modes of encoding in the brain and they're not exclusive. Much of the work on spatial mapping in hippocampal cells is on the topic of "place cells" which appear to encode with both rate and phase

[1] Dual phase and rate coding in hippocampal place cells: theoretical significance and relationship to entorhinal grid cells.
http://www.ncbi.nlm.nih.gov/pubmed/16145693

The wiki on neural coding has a good review of all the basic types of encoding:
http://en.wikipedia.org/wiki/Neural_coding

The main ones are rate coding, temporal coding, and population coding

 

[atyy]

A good introductory neurobiology book is available free at http://www.ncbi.nlm.nih.gov/books/NBK10799/?term=purves. It unfortunately cannot be freely browsed, although it can be searched.

The major topics in how a neuron spikes are the sodium and potassium channels in the Hodgkin-Huxley model.
http://www.ncbi.nlm.nih.gov/books/NBK10958/

These spikes are caused by strong enough input from other neurons, which are communicated by neurotransmitters and their receptors.
http://www.ncbi.nlm.nih.gov/books/NBK10807/
http://www.ncbi.nlm.nih.gov/books/NBK10960/
http://www.ncbi.nlm.nih.gov/books/NBK10977/

One example in which the strength of a synapse is changed that has been studied in detail is long term potentiation.
http://www.ncbi.nlm.nih.gov/books/NBK10878/

There is are also good introductory sections on the molecules underlying synaptic transmission in
http://www.ncbi.nlm.nih.gov/books/NBK21521/
http://www.ncbi.nlm.nih.gov/books/NBK26910/

 

[DiracPool]

The problem with selecting max 3 people gets worse every year, and it would have been surprising if this prize didn't labor under it. :w

I agree, but what really is the alternative? You have to select some number. The more you select, the more you attenuate the glory of the recipients, and that's kind of the whole idea of a prestigious prize, so I think we have to cut them some slack. I'm sure this job of theirs is not easy.

Personally, I think the Nobel should have gone to the technology makers of the recent revolution in neuroscientific data collection. Most of this revolves around non-invasive techniques related to the fMRI, but it goes well beyond that. A lot of it relates to data analysis techniques that abstract from these non-invasive raw data to really bring to light how large systems in the brain interact with one another. For instance, techniques in structural equation modeling, dynamic causal modeling, and especially Granger causality mapping.

I don't know if any of you remember, but I think it was Bush Sr. who declared the 1990's the "decade of the brain," and the neuroscience equivalents of Michio Kaku like Antonio Damasio and others were out touting the big advances going on, but nothing really happened in the 90's. The 2000's was really the decade of the brain in my opinion.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1011403/

The biggest gift we have been given over the last decade is the identification of the organization of "resting state networks." Thrillingly, our confidence in the fidelity of these non-invasive analyes has been further bolstered by corroborating incranial micro-electrode arrays studies on Granger causality measures from humans patients undergoing brain surgery.

The question is, how are you going to delegate these awards? I don't know, just run a search on any of the techniques I've listed above and you'll find hundreds of entries. However, I think if you're going to fall short giving credit where credit is due, at least do it in the right field.