Neuroscience is still the scientific Wild West: a wide open field. Computer modeling of brain structure/function is a way of systematically organizing into something resembling a comprehensible whole the vast quantities of seemingly unrelated data. "If you look at other branches of science, there are well developed methodologies and conceptual frameworks for thinking about chemical and physical problems," says Alex Protopapas, a dark, intense graduate student who spends much of his time glued to a computer screen creating simulations. "But neuroscience is like alchemy before it became chemistry. We don't have anything like a central dogma. Modeling might help us to develop some sort of methodology."

Above, neural circuit diagram of the piriform cortex. (Click to view larger.)

Simply stated, the kind of modeling done in Bower's lab is like reverse-engineering the brain by observing and analyzing computer simulations of cells and cell networks modeled on real brain structures such as the olfactory cortex and the cerebellum. These models are built from all the known anatomical and physiological properties of a particular brain structure --such as the purkinje cell network, or the electrosensory organ of the weakly electric fish -- then used to distill that data and load it up in the computer. At that point, with the aid of special simulation software, the information is used to create simulations of these brain regions in action. One of the biggest benefits of modeling is that this technique allows researchers to do trial runs of their theories without wasting precious laboratory resources and time.

"Modeling gives you a way of experimenting without actually having to do a real physical experiment," explains Dave Bilitch, one of the computer wizards who troubleshoots and oversees the operations of the lab's powerful hardware and software. "You can make a theory about what you think is going on, and then devise a realistic model. You might be able to observe something in the model that you can't see in the real cell, which can give you a lead to where you want to go next in your research. And if it doesn't work, you may be able to postulate what went wrong."

Theories about brain function coming out of these simulations are then tested for accuracy experimentally. "The best way to assure the most useful data for modeling is is to have experimentalists generate models," adds Bower, "and have modelers directly involved in experiments." That's the way the Bower lab is set-up, with a synergistic exchange between experimentalists, who plug in information from their research into neural simulation software programs like GENESIS, and modelers, who use their insights from simulations to design new experiments.

--Linda Marsa