Reverse Engineer the Human Brain
M.I.T. scientist Sebastian Seung describes the audacious plan to find the connectome--a map of every single neuron in the brain. Here, he says, is the secret of human identity
Tuesday, March 20, 2012 | 7 Scientific American....
Sebastian Seung Image: Cook: You argue in your book that neuroscience has a fundamental problem. What is the problem?
Seung: Most people are familiar with the regional approach to neuroscience: divide the brain into regions such as the "left brain" and "frontal lobe," and figure out what each region does. This approach has helped physicians interpret the symptoms of brain injuries, but at the same time has frustrating limitations. How do regions carry out their functions? Why do they malfunction in mental disorders? What happens to regions when we learn? We can never obtain satisfying answers to these questions if we consider regions as the elementary, indivisible units of the brain.
An obvious solution is to understand a region by subdividing it into neurons, and figure out how the neurons work together to perform the region's function. This neuronal approach has the potential to answer the big questions above, but so far has not succeeded. In fact, those who study regions sometimes criticize those who study neurons as too focused on minutiae.
Cook: What made you think that there is another way?
Seung: The neuronal approach is finally gathering steam because of technological innovations, especially in genetics and imaging. The nervous systems of animals can now be genetically engineered, allowing researchers to carry out much more precise and conclusive experiments. And there are powerful new methods of looking into the brain to see how neurons signal each other and how they are connected into networks. These developments make neuroscientists optimistic that we are finally going to understand the brain as a network of neurons.
Cook: What do you mean by the connectome?
Seung: A connectome is a map of a neural network. It is like one of those route maps you find in the back of airline magazines. Just replace each city with a neuron, and each route between cities by a connection between neurons. Keep in mind, though, that your brain contains about 100 billion neurons, so your connectome would never fit in the pages of a magazine.
Cook: Are there particular diseases which this research might help understand?
Seung: In brain diseases like Alzheimer's and Parkinson's, neurons degenerate and die. Autopsy reveals that something is visibly wrong with the brain. Yet for many mental disorders, such as autism and schizophrenia, a clear and consistent pathology of the brain has not been found. Why? Researchers have conjectured that the individual neurons are healthy, but they are connected with each other in an abnormal pattern. Unfortunately, such "miswirings" or "connectopathies" have remained merely hypothetical, because our technologies for mapping neural connections have been too primitive. Imagine what it was like to study infectious diseases before the microscope was invented. You could observe symptoms, but not the microbes that caused disease. Similarly, most mental disorders are still defined only by their symptoms. We need to uncover their causes in the brain, and the new field of connectomics will be important for that.
Cook: You have an interesting discussion of personal identity. Can you explain how this relates to the connectome?
Seung: I mentioned the hypothesis that certain mental disorders are due to abnormal neural connectivity. We could extend this idea to explain normal mental variation too. Perhaps minds differ because connectomes differ. You have probably heard people say things like, "Johnny's just that way. His brain is wired differently." I say it another way: "You are your connectome." We are the product of our genetic inheritance and our lifetime experiences. Genes have influenced your connectome in many ways--for example, by guiding how your neurons wired together during the development of your brain. Experiences have also modified your connectome, because connections are altered by the neural activity patterns that accompany experiences. To put it another way, your connectome is where nature meets nurture.
Cook: Mapping the connectome seems like an almost impossibly difficult challenge. Critics say that you will may never succeed, or that if you do it will take decades, and we can't put neuroscience on hold for that long.
Seung: Indeed, mapping an entire human connectome is one of the greatest technological challenges of all time. Just imaging all of a human brain with electron microscopes would be difficult enough. This would yield about one zettabyte of data, which roughly equals the world's current volume of digital content. Then analyzing the images to extract the connectome would be even more demanding. Yet I believe that we will eventually prevail. Success will not come with a sudden bang but rather through sustained growth over time. I imagine that the speed of mapping connectomes will double every year or two. If so, then it will become possible to map an entire human connectome within a few decades. There are similar success stories for other technologies. Computers have improved at this rate for the past half century. DNA sequencing has advanced similarly for the past forty years, and accelerated even further over the past decade.
That being said, such speculation about the far future is just for fun, and is actually beside the point. Even if we never succeed in mapping an entire human connectome, we will learn a tremendous amount by mapping connections in small chunks of human or animal brains. This trend has already begun. Exciting developments in connectomics are happening right now; we don't have to sit around waiting for the future.
Cook: Is there any way the research can be accelerated?
Seung: We invite the public to visit a web site called EyeWire, where you can help map the connectome of the retina, the sheet of neural tissue at the back of the eye. You don't need specialized training to participate, because EyeWire is like a virtual coloring book with pages that are images of the retina. (The images were taken with an electron microscope in the laboratory of our German collaborator, Winfried Denk.) Your task is to color in neurons, and you already know how to do this: just stay inside the boundaries. In this way, you will trace the "wires" of the retina, the branches of its neurons. This is the most laborious task required for mapping a connectome. (Another important task is identifying synapses, the tiny junctions at which neurons communicate with each other.)
EyeWire's coloring book is so vast that no single person could live long enough to manually color the neurons. We have sped up the process in two ways. First, artificial intelligence (AI) does most of the coloring automatically. You just have to guide the AI by a few mouse clicks here and there. Second, the coloring game is fun or even addictive to some people. Perhaps it's because the organic forms of neurons are mesmerizing. Or maybe it's because the game is challenging; at some image locations it can be difficult to decide whether there is a boundary between two neurons, i.e., whether to continue coloring or to stop. EyeWire users tend to improve with practice at such decisions, because they gradually learn from experience how neurons are shaped.
We are working to make EyeWire even more fun, in the hope of recruiting a large community of "citizen scientists." If each member of the community plays the coloring game a little, we can collectively map the retinal connectome. Community input to the site will also make the AI smarter, because the computer learns to emulate human judgments. This will accelerate the coloring process still further, until we will be ready to search for connectopathies. Philosophers love to ponder the question of whether the brain is complex enough to understand itself. Perhaps not, but maybe our billions of brains interacting with AI will be up to the task!
Cook: What made you think to turn to citizen science? Is it just a form of outreach, or do you really think it will end up having a significant impact on neuroscience?
Seung: We were impressed by the success of Galaxy Zoo in astronomy and FoldIt in molecular biology. Already a few years ago, we were thinking of creating EyeWire, but the required technologies were not yet available or widespread. When delivering 3D images to users, EyeWire works nicely with a 10 Mbps internet connection, a speed that has become common in households only recently. And EyeWire's interactive 3D graphics, rare for a web application, was implemented using WebGL. This standard is so new that it requires recent graphics hardware, can be tricky to configure in some older web browsers, and is unsupported by Internet Explorer. We hope that our users will understand that such annoyances come along with being an early adopter, but should disappear as the technologies mature.
EyeWire really excites me because of its potential for combining research, education, and outreach in a truly synergistic way. These activities are generally viewed as separate, and may even be seen as interfering with each other. Researchers may wish to spend more time on education and outreach, yet end up not doing so because they have to focus on research to remain competitive in their specialty. Likewise, educators may be too busy with teaching to do research. But EyeWire creates a situation in which important research goals hinge on the participation of citizen scientists. And learning science by actually doing science may turn out to be more effective than traditional educational methods, or at least complement them nicely.
Cook: Whether the public is helping or not, mapping the connectome will only provide the structure of the neural network, not the signals that the neurons are actually sending. Aren't you just setting yourself up for another, even more daunting project?
Seung: Using new methods of light microscopy, neurophysiologists are now able to image the signals of hundreds or even thousands of individual neurons at the same time, in the brains of living animals. (Compared to microscopy, MRI has the advantage of being applicable to living human brains but blurs 100,000 neurons into a single pixel.) Such studies of neural activity can be followed by electron microscopy to map the connections of the same neurons. Imagine knowing the activity and connectivity of all the neurons in a small chunk of brain. This capability is finally within reach, and is bound to revolutionize neuroscience.
From Scientific American
Are you a scientist who specializes in neuroscience, cognitive science, or psychology? And have you read a recent peer-reviewed paper that you would like to write about? Please send suggestions to Mind Matters editor Gareth Cook, a Pulitzer prize-winning journalist at the Boston Globe. He can be reached at garethideas AT gmail.com or Twitter @garethideas.
No hay comentarios:
Publicar un comentario