Sebastian Seung: We Are Our Connectomes | TED Talk

[rhody]

Latest http://www.ted.com/talks/sebastian_..._campaign=newsletter_weekly&utm_medium=email" September 2010

In the brain, neurons are connected into a complex network. Sebastian Seung and his lab at MIT are inventing technologies for identifying and describing the connectome, the totality of connections between the brain's neurons -- think of it as the wiring diagram of the brain. We possesses our entire genome at birth, but things like memories are not "stored" in the genome; they are acquired through life and accumulated in the brain. Seung's hypothesis is that "we are our connectome," that the connections among neurons is where memories and experiences get stored.

Seung and his collaborators, including Winfried Denk at the Max Planck Institute and Jeff Lichtman at Harvard University, are working on a plan to thin-slice a brain (probably starting with a mouse brain) and trace, from slice to slice, each neural pathway, exposing the wiring diagram of the brain and creating a powerful new way to visualize the workings of the mind. They're not the first to attempt something like this -- Sydney Brenner won a Nobel for mapping all the 7,000 connections in the nervous system of a tiny worm, C. elegans. But that took his team a dozen years, and the worm only had 302 nerve cells. One of Seung's breakthroughs is in using advanced imagining and AI to handle the crushing amount of data that a mouse brain will yield and turn it into richly visual maps that show the passageways of thought and sensation.

• 4:40 Slices of mouse brain: magnified 100,000 times to see branches of individual neurons
• 5:00 Images stacked provide a three dimensional image of the brain
• 6:15 Every neuron is given a unique color
• 6:45 How are brains of men and women different, (woah, here we go...)
• 8:10 Cube of neurons = 6 microns on each side, very very small
• 8:30 Sebastian, you should just give up, neuroscience is hopeless
• 9:20 Curiosity + wonder sometimes leads to despair
• 10:05 Finding human connectome greatest challenges of all time, it will takes generations
• 10:30 Modest goal: to study partial connectomes of small chunks of mouse and human brain
• 11:50 What Sebastian calls neural activity I believe is also referred to as neural plasticity
  because he says neural activity can cause your connections to change (same as neural plasticity), your experiences can change your connectome
• 14:00 How to test the hypothesis, I am my connectome: attempt to read out memories from connectomes: long termporal memories of sequences of movements of a classical pianist
• 16:15 May allow researchers to see mental disorders in the brain, autism, etc...

Summary: I found the presentation interesting, a little preachy, a bit over the top at times, however, you could sense Sebastian's enthusiasm for his research, that made it worth watching.

Comment: I know there are a few folks on PF who follow this subject closely and may be or are on a path for research in neurology, it would be nice to hear from them.

I suspect, however, taking all of what we know about how the Universe operates at present, we still need to account for the unpredictability of QM in a truly robust theory of connectomes. That in an of itself adds another whole new level of complexity (and for me at least, wonder) to the study of neurology. But... first things first, if we can test and prove the macro world of chemical and electrical activity in the brain (I have personal doubts, however) to some level of repeatability it will encourage more research with connectomes, if not, then whatever works and replaces it will.

Rhody...

 

[apeiron]

I suspect, however, taking all of what we know about how the Universe operates at present, we still need to account for the unpredictability of QM in a truly robust theory of connectomes. That in an of itself adds another whole new level of complexity (and for me at least, wonder) to the study of neurology.

No, I think you can safely rule out any need for QM level precision. Indeed, we know enough about neurology to say that even around about the level of individual synapses, their contribution to the "connectome" is statistical rather than deterministic. Indeed, you can blast quite a lot of holes in the connectome (strokes, alzheimers, etc) and the brain still hangs together. It has a lot of top-down capacity to fill in, to cope with noisy inputs.

But this idea of slicing the brain and reconstucting a 3D map of its connections has been around a while - long enough that I can't remember who was first telling me about their hopes of getting a project off the ground.

It'd certainly be great if it could be done - and there was a way to datamine for some simple models of canonical neuroanatomic principles. So just doing the mouse cortex would be enough of a start here.

However the project has its natural limits. A map of connections does not tell you which connections are being enhanced, which gated, when actually doing anything. And of course, that is fundamentally important.

Although perhaps it is less of a biggie than making a link between genes and the processes they control. With brains, visible structure does tell a lot about function.

 

[Ygggdrasil]

As apeiron said, there are limits to how much information a connectome can tell us. For example, we have known the C. elegans connectome for quite some time, yet we still don't have a good idea of how neuronal circuits work in this organism. Likely, some combination of a connectome along with many, many functional studies are necessary. One tool that will likely help is the ability to manipulate the activities of individual neurons using optogenetics (e.g. activating/inactivating specific neurons using light-sensitive ion channels developed originally by Deisseroth and Boyden).

Another difficulty is that most connectomics efforts create their 3D reconstructions using electron microscopy because standard light microscopy does not have sufficient resolution to trace neurons and identify synapses. While EM has the advantage of increased spatial resolution, it lacks the ability to easily identify different molecules in cells (methods do exist, such as using antibodies linked to gold particles that are visible by EM, but these methods do not work so well). In contrast, by fluorescently labeling different cellular components (for example with a fluorescent protein), light microscopy can provide information about the identity and function of a synapse (e.g. which type of neurotransmitters act at the synapse, whether markers of different types of plasticity are present). Therefore, one promising area here is to could combine the light microscopy's ability to detect specific molecules with high resolution imaging. One method would be to image the same sample with both light microscopy and EM and overlay the information (e.g. the "array tomography" approach developed by Steven Smith). Another means would be to use one of the newer superresolution microscopy approaches (e.g. PALM, FPALM, STORM, STED) which are light microscopy techniques that give much higher resolution than conventional light microscopy.

 

[Ruslan_Sharipov]

Sydney Brenner won a Nobel for mapping all the 7,000 connections in the nervous system of a tiny worm, C. elegans. But that took his team a dozen years, and the worm only had 302 nerve cells.

It would be best for the beginning not to slice a mouse brain, but to pin tiny wires to each of 7,000 connections of Caenorhabditis elegans and, using a computer, get a time graph of electrical activity in these connections. The worm should be kept alive during this procedure.

P.S. Please, where can I get the connection diagram obtained by Sydney Brenner?

 

[Ygggdrasil]

P.S. Please, where can I get the connection diagram obtained by Sydney Brenner?

The C. elegans connectome can be viewed at http://www.wormatlas.org/

 

[Phrak]

"Seung's hypothesis is that "we are our connectome," that the connections among neurons is where memories and experiences get stored."

This is hardly a new idea, right? It is something I presumed true 30 years ago.

 

[Kmenex]

What about the hemodynamics of the brain? that is not very well understood either..

 

[Phrak]

I'm sure hemodynamics plays a role in memory, as does feedback from heart and gut. Not so much what we remember but what memories are triggered--that is, which system of neurons are active. However, neuron loops through the rest of the body cannot be left out of synaptic patterns without altering a memory. And memory is surely more involved than synaptic potential alone, involving a host of chemical interactions.

We people think we have accurate memories of the past. This is, for the most part, nonsense, but really the best rational way to proceed. It wouldn't do any good to go on the assumption that everything you think you know is false. Memory is colored and rewritten with each access by the conditions that invoked it.