September 1999

Getting the Big Picture: Modeling Anatomically Correct Neurons By Emily Yaghmour He has a Ph.D. from the most prestigious university in Italy, and he has brought in more than half a million dollars in research money in the two years since he joined the Krasnow Institute for Advanced Study. Not bad for a 29-year-old. Giorgio Ascoli, who also serves as a visiting professor in the Department of Psychology, grew up in Milan, Italy, the son of a physics professor and a high-school teacher. After high school, he was accepted into the chemistry program of the Scuola Normale Superiore university in Pisa, a school so prestigious and selective that only four slots were open to chemistry students the year Ascoli was accepted. Founded by Napoleon, the school has a one-to-one ratio of faculty to students, and the undergraduate student body consists of but 200 students. The students are treated like royalty. Their beds are made each morning, and their clothes are washed and ironed for them. But don't be jealous. The curriculum is enormously demanding and requires everyone to maintain a 4.0 GPA. One-third of the freshman class flunks out before the second year. "The first year was hell," Ascoli remembers. But to his surprise, he made it. "It was an enormously stimulating environment--if you could survive it," he smiles. Not only did he make it into the selective undergraduate program, but he made it into the even more selective Ph.D. program, which had only one opening the year he applied. Through this program, he landed a doctoral research opportunity at the National Institutes of Health (NIH) in Bethesda, Md., where he worked under the supervision of James L. Olds, a neuroscientist and affiliate of the Krasnow Institute. Olds, who is now director of Krasnow, was impressed with the ingenuity Ascoli brought to a particularly formidable research problem Olds had asked him to work on. "It left me with the impression that Giorgio was not your typical student. He was amazingly brilliant." When Ascoli graduated with his Ph.D., he turned down a handsome offer for a postdoctoral fellowship at one of the best Alzheimer's labs in the country to come to Krasnow. Krasnow is unusual in that it encourages its researchers to explore the mind-body connection. This particularly appealed to Ascoli, who had grown frustrated with conducting lab experiments, the results of which were important in illuminating certain aspects of neuroscience but revealed little about the overall operation of the brain. "Not that details aren't important," says Ascoli, but what interests him is the broader picture. Many scientists shy away from this broader picture, he says. At many institutions, "if you tell a colleague you're interested in studying human consciousness, he will tell you, 'Shh, don't say that until you get tenure.'" But at Krasnow, not only can you freely admit it, this line of investigation is encouraged, he says. Shortly after Ascoli joined Krasnow in 1997, he put together a research proposal that recently won a huge grant from NIH. The proposal is to build virtual neural networks, using a new approach that he and others on his research team devised. The correct anatomical description of a single neuron can take thousands of lines of a data file. Since a human brain contains many trillions (one trillion is one million million) of neurons, modeling even a tiny portion of neural circuitry takes not only an immense amount of computer memory but also an enormous amount of time. Ascoli found a way around this. Even though the brain has trillions of neurons, it has only a few thousand different types or "morphological classes" of neurons. Scientists have long known that neurons have certain anatomical characteristics that differ from class to class. Like snowflakes, no two neurons are identical, but the anatomical characteristics of neurons within a single class share the same parameters--for example, the diameters of the trunks of neurons in the same class are statistically within a certain size, larger than x and smaller than y. Ascoli used an algorithm-a step-by-step procedure-that incorporates random numbers to generate virtual neurons of a certain class. The numbers are selected randomly from the experimental parameters of a given anatomical characteristic. For example, since scientists know that neurons in the same class have trunk diameters that are larger than x and smaller than y, the algorithm randomly takes numbers between x and y and uses them to produce neurons with diameters in the given statistical range. It does the same thing for several other anatomical characteristics. In this way, Ascoli and his colleagues are able to computationally generate millions of anatomically correct virtual neurons using a fraction of the time and energy it takes to model a single neuron the old way. Ascoli is, of course, not the first scientist to design neural networks. Others do it also, but most of them don't design anatomically correct neuron models, because it takes so much time. Their models are more abstract, using sticks and balls, or zeros and ones, to represent neurons. Their focus is on how neurons produce certain behaviors--for example, how neurons enable fingers to type words on a keyboard. But isn't this the real object of brain research--to determine how the brain lets us do what we do? Yes, but scientists know that anatomical characteristics have a lot to do with the behaviors that neurons invoke. By producing anatomically correct neurons and employing tools that simulate the activity of the neurons, researchers can better discover how the anatomy of neurons influences behavior. "My approach is both humble and grand," explains Ascoli--humble because it returns to the concrete reality of structure, grand because no one else has done it and because it is capable of producing, theoretically, an anatomically correct (down to the finest detail) virtual brain. The expectations for the research are both humble and grand as well: grand because a truly realistic model can help scientists understand what's going on, so that they can better treat diseases and disorders of the brain; humble because it can serve as a teaching tool for undergraduates interested in neuroscience. "I know it certainly helped me," says Ascoli. A humble statement coming from a young man with such grand accomplishments.