Connectomics- mapping brain connections

[Dima]

There is a debate as to whether mapping brain connections on the microscale with electron microscopy is a worthy undertaking.

 

I think it will be awesome and comparable to sequencing a genome. Once we map brain connections in diseased brains we can compare them to normal, compare mice to monkeys and eventually to humans in diseased and normal states and young vs old.

 

Does anyone think for example: a project to map the brain of a an ape is worth pouring billions of dollars into; why/not ?

 

Requirements:

1. Faster methods of imaging brain slices with an electron microscope

2. Machine learning computers to trace out all the connections at synapses from those images

3. Capacity to store potentially many zetabytes of imaging and analysis data

 

A critical look at connectomics:

http://www.nature.com/neuro/journal/v13/n12/full/nn1210-1441.html

[EdEarl]

Suppose you had been around and Ben Franklin had asked you, "Do you think I can learn anything if I fly a kite in a thunderstorm?" Based on knowledge of the times, you know nothing about electricity, and you know lightening sometimes strikes and burns down buildings. What answer would you give?

1) You will revolutionize science.

2) I don't know.

[CharonY]

My feel is that it could be useful, but far from being revolutionary (i.e. pretty much the same as the author of the linked article). If we use the same analogy of genomics, it does provide a lot of useful information, at the same time we cannot really keep up with interpreting the same. The reason is that lack a strong theoretical framework in which we can interpret the data. Often times we see differences in omics data (e.g. diseases vs healthy, stress vs unstressed etc.), get a set of changes and have no clue how that translates into biological outcomes. The thing we learned from the omics revolution is that massive measurements will give us massive amounts of data, but by itself will not help us interpret that data in a biologically meaningful way.

I have to add that this is not my specialty, but the articles I have read are very similar to those about omics in the late 90s. A lot of enthusiasm and a strong focus on the technical aspects, however not enough insights in what to do with the data.

[Ringer]

I think there are two things to take into account when looking at any massive undertaking in science, but brain research tends to be the focus right now. One is that we never know what the effects will be when we get the information, but we can assume that getting massive amounts of new information will massively change the way research can and will be done. The revolution will not be seen for a long time from a layman's perspective, but for the research it will change a lot. This means the connectome projects will probably not directly change anything for most people, but it can be used by researchers to test ideas that can revolutionize things for normal people.

 

Second, and most important for most people, almost all science coverage and communication cannot rationalize the spending in ways most people can understand. Not that people are stupid, it's just they don't understand what science entails. Projects such as these help scientific revolutions the same way other projects do, by small increments; they just cost more, but are somewhat necessary to go in the direction research wants to go. To explain it like this would massively undermine support for the spending because people want immediate results for their investments. So you have overestimates of how revolutionary that data will be.

 

tl:dr version: Sometimes scientific endeavors are overhyped, but they are still extremely important in the long run

[CharonY]

Or in some cases they just show what does not work, which by itself can also be rather valuable (though often not highlighted for obvious reasons).

However, as a reviewer I would gladly throw the same money on a more limited study that provides approaches to make something with the data. Unfortunately sometimes the big ones are an easier sell as people oftentimes still prefer to fund the next big thing (especially if bigshots are involved) rather than going for slower but incremental gain of knowledge. Even if the big thing eventually amounts to less in the end.

 

Though this arguably due to the the semi-politization of science funding.

Edited July 8, 2013 by CharonY

[Dima]

Suppose you had been around and Ben Franklin had asked you, "Do you think I can learn anything if I fly a kite in a thunderstorm?" Based on knowledge of the times, you know nothing about electricity, and you know lightening sometimes strikes and burns down buildings. What answer would you give?

1) You will revolutionize science.

2) I don't know.

Yes

Or in some cases they just show what does not work, which by itself can also be rather valuable (though often not highlighted for obvious reasons).

However, as a reviewer I would gladly throw the same money on a more limited study that provides approaches to make something with the data. Unfortunately sometimes the big ones are an easier sell as people oftentimes still prefer to fund the next big thing (especially if bigshots are involved) rather than going for slower but incremental gain of knowledge. Even if the big thing eventually amounts to less in the end.

 

Though this arguably due to the the semi-politization of science funding.

Why do you think that we can't "make something" with the data? The data will show how brain connections differ in pathologies.

[CharonY]

I am not saying that we cannot do anything with it. What I am saying that unless someone comes up with a good way to interpret the data we will remain descriptive in nature.

E.g. we may see differences between different states, but if it is anything like omics data, we will have at least two challenges. The first is that most like the differences will be subject to quite a bit of biological noise. I.e. we are likely not getting clear-cut and reproducible difference between a "baseline connectome" (which is more or less arbitrarily what we define as normal) and pathologic states. The exceptions could be in the extremes where major differences start happening. But that does not necessarily teach us a lot of how the brain normally works.

That leads to the second challenge. How do we translate these descriptive information into something that is of physiological relevance?

 

Of course, there may be approaches that may be viable and since I am not an active researcher in this field it may be just my ignorance. However, articles are these posted in OP sound awfully the same as the issues we are facing in the omics field (which happens to fall into my area of expertise).

Edited July 8, 2013 by CharonY

[Dima]

I am not saying that we cannot do anything with it. What I am saying that unless someone comes up with a good way to interpret the data we will remain descriptive in nature.

E.g. we may see differences between different states, but if it is anything like omics data, we will have at least two challenges. The first is that most like the differences will be subject to quite a bit of biological noise. I.e. we are likely not getting clear-cut and reproducible difference between a "baseline connectome" (which is more or less arbitrarily what we define as normal) and pathologic states. The exceptions could be in the extremes where major differences start happening. But that does not necessarily teach us a lot of how the brain normally works.

That leads to the second challenge. How do we translate these descriptive information into something that is of physiological relevance?

 

Of course, there may be approaches that may be viable and since I am not an active researcher in this field it may be just my ignorance. However, articles are these posted in OP sound awfully the same as the issues we are facing in the omics field (which happens to fall into my area of expertise).

 

Sure. Also, I think that all the data will not only be for connectomics aka mapping brain connections per say; you can extrapolate a TON of other science from stacks of EM micrographs, so if all this data is shared with other researchers, we can:

 

1. map brain connections

2. see what other changes occur compared to control. For ex: changes in volumetric parameters; organelle size, synapse density. Tweak some methods and youll also see protein localization and identify cells.

3. Reconstruct all that to make awesome 3D models

 

This is a lot work and currently the problem is in computation.

 

As for the biological noise problem, I agree we haven't a clue about how all the different cells of the brain work in concert; I also think a sure-fire way to create even more biological noise is to approach the problem from a biochemical or molecular perspective: with all the different cells in the brain it will be even more difficult to put all that data together as a whole once we have all the molecular pieces and most likely, they will not coincide with each other.

 

Electron microscopy, on the other hand, provides data that you can see and use readily.