In June last year, IBM and the Brain Mind Institute at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland announced a plan to create a digital 3D replica of the brain. Named after the IBM Blue Gene supercomputer it relies on, the Blue Brain Project has started modelling--in every detail--the cellular infrastructure and electrophysiological interactions within the cerebral neocortex, which represents about 80% of the brain and is believed to house cognitive functions such as language and conscious thought.
"The main problem in computational neuroscience is that theoreticians [who] do not have a profound knowledge of neuroscience build models of the brain," writes Henry Markram, founder of the Brain Mind Institute and director of the Blue Brain Project, in an e-mail. Current models "may capture some elements of biological realism, but are generally far from biological." What the field needs, he says, is "computational neuroscientists [who are] willing to work closely with the neuroscientists to follow faithfully and learn from the biology."
Twenty-nine-year-old Felix Schürmann, a German physicist on the Blue Brain Project, has taken up the challenge. When he first joined the project as a postdoc, "His task was on a specific development and generally to assist me in the management," says Markram. But "The progress in [his] first week was so dramatic that I allowed him more autonomy to manage the project." One year in, Schürmann oversees all aspects of the project, managing an international team of 35 scientists. "[Felix] is absolutely outstanding," says Markram. "Within weeks he had the project running at full speed, so much so that we will complete the cellular-level model by the end of the year."
Getting a full view of computer science
"I was intrigued by computing from the start," says Schürmann, who got his first contact with computers through volunteer classes as a schoolboy. When the young Schürmann tried his hand at programming, it was quickly apparent that he had a talent for computer science. In 1995, he became one of the winners of the German National Computer Science Competition, which involved solving computer problems, devising running demonstrations of the solutions, and surviving a 3-day final round packed with teamwork, oral presentations, and interviews with computer science professors. Based on this success he was offered a university scholarship from the German Academic Foundation a year later.
Schürmann´s interest in computing quickly reached beyond programming. "I wanted to understand the [different] aspects of it," he says. At university he studied physics--not computer science--"because I wanted to understand how computers work but also the constraints behind it." Schürmann earned minors in computer science and mathematics in addition to his physics major, obtaining his Vordiplom from the Ruprecht-Karls-Universität in Heidelberg in 1999.
His degree in hand, Schürmann left Germany for the United States where, he felt, computer science could be combined more easily with physics or industrial research. He did an M.Sc. in physics at the State University of New York (SUNY) at Buffalo, writing a thesis on interactive quantum computation--"the idea … that you can use the properties of … quantum mechanics to build different kinds of computers." He funded his stay in the United States with a Fulbright Fellowship.
Schürmann returned to Heidelberg for a Ph.D. on hardware-implemented neural networks, joining the "Electronic Vision(s)" group in the physics department, which he had become acquainted with while he was still an undergraduate. He wanted to work with them because "they are physicists and develop microchips" containing analogue integrated circuits rather than the less complex digital microchips. Schürmann already had seen the soft side of computing; "This Ph.D. gave me a chance to find out about the second half: the hardware. It was the completion of my view of how computers work."
Schürmann’s Ph.D. project was to implement an artificial neural network, an assembly of interconnected artificial nodes that process information according to a mathematical or computational model. "We still don’t understand how the brain works" and artificial neural networks, Schürmann says, are likely to be an important part of that understanding when it emerges. As his group entered biology-focused collaborations, Schürmann got a first taste of the challenges of merging the two fields.
Probing the complexity of the brain
One year before the end of his Ph.D.--which he received in 2005, magna cum laude--Schürmann met Markram, his current boss, as his group began to collaborate with the Brain Mind Institute. "I was impressed by [Markram's] openness and understanding of technological solutions, even though he is a biologist," says Schürmann. "He really combines biology and technology." He heard about Markram’s plans for Blue Brain Project and wanted to get involved.
The aim of the Blue Brain Project is to build a replica of a neocortical column, the basic functional unit that makes up the cerebral neocortex while encompassing most of the neocortex's cellular diversity. "If you are an experimental biologist, in your experiments you see an amazing variety of cell types. … The typical modelling approach ... doesn’t give the answer to the experimentalist who wants to understand this diversity," Schürmann explains. This diversity can only be achieved, he says, by incorporating experimental neuroscience data into very detailed computer simulations that "behave indistinguishably from the experiment." The Project builds on the efforts of the Brain Mind Institute, which has been accumulating empirical data on the microarchitecture of the neocortex for a decade.
In this, his first postdoc, Schürmann leads a team of 23 computer engineers and computational neuroscientists and 12 electrophysiologists based in Lausanne, Israel, and the United States. As project manager, Schürmann leads the project jointly with Markram and two scientific team managers--Idan Segev from the Hebrew University of Jerusalem in Israel, a pioneer of modelling complex neurons, and Philip Goodman, from the University of Nevada, Reno, in the United States, a long-time collaborator on large-scale modelling and advanced bioinformatics. The first task for Schürmann was consolidating the neuroscience data. "Biological data so far have been taken around the world in many labs," he says, each using its own standards. "To use biological data in a quantitative way, you really have to understand the data and control the way you got these data." Otherwise, he says, it isn't possible to put them into a conceptual framework where they can be built into detailed simulations.
One year into the project, a simulation of a neocortical column is already in place. "We built and simulated 10,000 compartmental neurons with over 30 millions dynamic synapses and are fine tuning the biological parameters. This is several orders of magnitude larger and more detailed than any previous attempt," says Markram. The project is now entering the calibration phase, which means making sure that every piece of the simulation is validated by experimentation.
In the long run, the new simulations may be used as a tool to better understand brain functions and to "find out what processes in this intricate network are the fragile ones" that lead to diseases such as epilepsy. It may eventually be possible, Schürmann says, to use the simulations in pharmacological tests. Advances are also likely on the computational side. "The brain is the computer in the world that does the most fabulous things. Finding how computing can be done differently will change our technological environment."
Combining computer science and neuroscience
According to Markram, people who want to work in this field need all the basic computer science tools, such as a perfect fluency in C/C++ and Matlab and knowledge of Web technologies. Students should be able to demonstrate expertise in constructing software models and implementing and testing algorithms. Knowledge of more advanced computer science tools, like visualisation technologies, is an asset. But "The next most important requirement is math and physics, as this allows the easy and efficient implementation of complex algorithms," Markram says.
A neuroscience background is also very important. "The greater the knowledge, the more autonomous the student can become," says Markram. "Experience in a neuroscience wet lab is a great accelerator." Another key ingredient is a certain breadth of perspective. "The main difficulty for all computational neuroscientists is the tendency to focus on isolated problems. The real challenge is to place even a single molecule in[to] the context of a behaving organism. To do so, computational neuroscientists have to progressively acquire a deeper knowledge of neuroscience from the genetic to the behavioural levels. It is essentially a second Ph.D. in neuroscience."
Schürmann matched up well with these requirements. "Felix came from a background of physics and computer engineering. His Ph.D. was in building hardware neural models, which is perhaps the most challenging task possible in computational neuroscience. He therefore had all the skills and experiences necessary for the position. The most important quality was … that he recognised that while building simple models is important … we can only answer the deeper questions about the complexity of the brain by modelling the complexity." Schürmann also came across as an excellent team-player--a characteristic that's essential to big, integrative projects like Blue Brain--and quickly learned more about neuroscience.
As for Schürmann's managerial role, neither Schürmann nor Markram expected it. "I did not really know that Felix would be able to lead the project," says Markram. "He seemed highly organised, [and] very quickly assimilated the project.... He asked very important questions concerning the feasibility of the project, but did not do so sceptically, which indicated to me that he was not afraid of entering a grand challenge; he was already preparing to take on the challenge" when he arrived.
"The Blue Brain Project will expand in 2008 to become a multinational project of enormous scale," says Markram. "Hopefully, Felix will take on this next leap in the challenge" and manage this expansion of the project, which will start in 2007. "After Blue Brain, I think Felix will take on only bigger-vision projects. It would be tough to step back."
Elisabeth Pain is contributing editor for South and West Europe.
Comments, suggestions? Please send your feedback to our editor.