Neurobiology and mathematics are two different worlds with two completely different languages. However, the two disciplines have come together in research into the brain. While neurologists hope to unveil the brain’s secrets through measurements, mathematicians are using formulas and theoretical models to describe brain functions. Stefan Rotter from the Bernstein Centre for Computational Neuroscience (BCCN) in Freiburg speaks the language of both disciplines, as he has both a mathematics and a physics background as well as having worked at a number of biology institutes. We ask the question: what is it like to wear so many different hats?
The human brain consists of many billions of nerve cells. Each of these neurones establishes thousands of contacts with neighbouring cells in order to exchange electrical signals. The consequence of this apparent chaos is our way of thinking, feeling and remembering. But how does the concrete structure of a neuronal network determine its function? How are the spatial and temporal patterns of electrical activity that encode this information produced? “These questions can only be answered when many different disciplines work closely together,” said Professor Dr. Stefan Rotter, Director of the BCCN in Freiburg and computational neuroscientist, who plays the role of mediator between several subdisciplines of brain research.
Rotter was born in 1961 in Landshut and brought up in the small city of Burglengenfeld in the Upper Palatinate. Like many other young people in eastern Bavaria, he wanted to escape the countryside and discover the world. But why did he chose to study mathematics and physics? "I chose these subjects partly because I enjoyed them at school and partly because I got good results without having to dedicate too much time to my work," said Rotter. During his advanced studies in Regensburg, Rotter soon realised that scientific mathematics had a lot more to offer than he had thought at school. He was fascinated by abstract objects. As a student, he spent a few months on the East Coast of the United States before returning to Hamburg where he graduated in 1989 with a thesis on associated algebras. "I was really fascinated by the topic, but I did not like the way mathematicians worked," said Rotter. "Many mathematicians are real mavericks who only work alone."
During a visit to the Max Planck Institute for Biological Cybernetics in Tübingen, Rotter met his future supervisor, the neuroanatomist Valentino Braitenberg. He was fascinated by the way the biologists, physicists and mathematicians in Braitenberg’s team worked together and discussed issues. Rotter decided to go to Tübingen and study the brain. As a mathematician, his work was to ‘mathematise’ the neuroanatomical and neurophysiological aspects of the neocortex discovered by his experimental colleagues. He was seeking answers to questions such as: why are the cells in a particular part of the brain connected in this way and not that? Which cell communicates with which cell and why? What does this mean for the function of the network?Rotter’s mathematical know-how was (and still is) very helpful for him because traditional physics methods are not always sufficient to describe neurones and networks. Neocortical neurones communicate by way of action potentials, short and predictable alterations of the membrane voltage. Information is not encoded gradually, but according to the all-or-nothing principle. “Models of such neurones need to be based on discrete variables,” said Rotter. “That is why I decided to look into the theory of stochastic point processes. This is very complicated mathematics, which at the time, had not been systematically used by brain researchers for the analysis of networks.
It was very unusual for a mathematician to want to do a doctorate in neurobiology, and even more unusual was the intention to do this at an extramural research institution. Therefore, Rotter was not only caught between two research themes, but also two different research practices. After several years, during which he often had to overcome bureaucratic red tape, he finally received his PhD in physics in 1994. He continued his scientific career as a postdoc at the Max Planck Institute for Developmental Biology in Tübingen, which was equipped with the very latest computers. The modern equipment enabled Rotter to optimise the modelling of theoretical neuronal networks using computers. He was able to implement hypotheses on new experimental findings in neuronal networks, test his ideas and come up with predictions for potential new experiments. During this time, the discipline of computational neuroscience was rapidly emerging. Scientists who not only had an understanding of the biology of the brain, but who were also acquainted with the complex mathematics required for theoretical analyses and complex data processing were in great demand.
In 1996, Rotter moved on to the Institute of Biology III at the University of Freiburg to become one of Professor Dr. Ad Aertsen’s assistants and contributed to the establishment of the Department of Neurobiology and Biophysics as the head of a group dealing with theoretical neurobiology and biophysics. The close exchange with experimental biologists had now become an integral part of his work. In 2002, Rotter moved on to the privately financed Institute for Frontier Areas for Psychology and Mental Health in Freiburg, but remained in close contact with the Department of Neurobiology and Biophysics. In 2003, he habilitated in the areas of neurobiology and biophysics. In 2004, he co-founded the BCCN and became its director in April 2008 at the same time as receiving a full professorship from the University of Freiburg.
Rotter and his team are still working on the neuronal networks of the neocortex. The scientists are able to simulate the activity of up to 500,000 individual cells in their models. This is achieved by connecting several dozen standard computers together. One particular aspect of research involves the investigation of systems that alter their connections as a result of previous activities. At some stage in the future, these theoretical experiments may be able to help scientists to understand how learning functions. Other approaches also take into account growth- or age-related structural changes in the networks. “The structures, which consist of many billions of connections, are very complicated per se,” said Rotter adding, “since all these systems change over time, it is easy to see how the analysis of the simulations could become difficult.”The simulations result in vast quantities of data. Therefore, Rotter and his team increasingly have to use statistical methods for the analysis of their data. How can relevant aspects be sifted out of the chaos of signals? How can the vast quantities of data be analysed quantitatively and systematically? The new methods benefit Rotter and his colleagues, who use experimental approaches in order to gain insights into cerebral function, including electrophysiological recordings. “We are trying to bring together theory and experimentation,” said Rotter. “And this is maybe the best definition of ‘computational neuroscience’. In their efforts to be able to understand the brain in all its complexity, scientists like Stefan Rotter need to have a greater understanding of a number of disciplines.