Biologists tend to be a reductionist lot. They see a system or a pathway and pull it apart to try to understand how it works. The problem for biological systems in general and the nervous system in particular, is that once you've pulled it apart, it's very difficult to put it back together.
"How do you know if you really understand any system?" asks computational neuroscientist Terry Sejnowski, whose lab is focused on understanding the principles linking brain mechanisms to behavior. "You have to put the system back together to see if it works. With a biochemical pathway, you reconstitute the system from the individual parts you've identified. In neuroscience, you reconstitute a system by running computer simulations of networks of neurons to see if you can re-create the behavior."
Sejnowski sees the challenges of reconstructing neuronal behavior in computer models as a one big loop of experimentation: develop theoretical models of neuronal behavior, test those models, and refine them—just to start all over again.
A physicist by training, Sejnowski finds the process of developing and testing theoretical models to be second nature. "I was fully intending to go into physics," says Sejnowski, who sought to study gravitational radiation—the as-yet-undetected radiation emitted when massive bodies accelerate. However, experiments to detect this last unknown in Einstein's theory of general relativity are enormous collaborative efforts requiring hundreds of millions of dollars. "I realized that I might go my entire career without having any of my ideas tested," he says.
Searching for a field that would allow him to develop theories and test them through experimentation, Sejnowski decided to complete postdoctoral training in neurobiology at Harvard Medical School, studying synaptic physiology.
"What could be more exciting than trying to understand how the brain works?" asks Sejnowski. "I was really curious because there are all these big questions. Why do we sleep? That's still a mystery. What is consciousness? We don't know. "
Part of his fascination came from the fact that "when you look at the brain you realize this is not the way an engineer would design the system. The brain is redundant, massively parallel, and regenerative," he says. "Neurons dedicated to one activity can be put to new uses," a characteristic known as plasticity.
As he pursues the study of sleep, memory consolidation, visual processing, and sensorimotor coordination, Sejnowski incorporates experimental findings into computer simulations to repeatedly test his theoretical models.
One of the biggest obstacles to testing those models has been the need to develop high-powered computational tools. The state-of-the-art MCell cellular simulation program, a collaboration between the Pittsburgh Supercomputing Center and the Salk Institute (with support from the National Institutes of Health, HHMI, and the National Science Foundation), took more than 15 years to develop. MCell allows scientists to account for every single molecule and protein inside and outside a cell and document their activities on microsecond timescales. Using the program, scientists can simulate how neurons use those molecules and proteins to communicate with each other.
"We're trying to follow information as it goes into the brain and put all the pieces together so that we can really understand what happens, for example, when you see something, " says Sejnowski, adding that sometimes the process of following that information can fundamentally change your view of how a system works.
All scientists studying the brain's network of neurons have a conceptual framework for pursuing their experiments. For example, for decades the assumption was that in the visual cortex—the folded section of the brain responsible for vision—the most important information from neurons was the total number of times the neurons fired.
Sejnowski, however, maintains that the pattern of nerve impulses is also important for understanding brain function. Nerve impulses—called spikes—are pulses of voltage that travel along the axons of neurons. By analyzing the timing of these spikes, neuroscientists have found that when a single synapse is stimulated a few milliseconds just before or just after a spike in the neuron, the strength of the connection to the neuron increases or decreases, respectively. In effect, this regulates the degree to which neurons fire together, or synchronize their firing patterns.
"That synchronization is regulated by the brain somehow," says Sejnowski. Spike timing and the synchronous firing of large numbers of neurons in the visual cortex may be used to enhance the quality of sensory inputs. He says that synchronization in the visual cortex is regulated by the act of paying attention. As a result, sensory input may have more or less impact, depending on whether you are paying attention.
Sejnowski says the role of attention on visual perception is apparent whenever we fail to see something because we don't expect to see it. "You can be trained to be observant. Ironically, we know more about how the brain learns than we know about how the brain pays attention," he says.
By probing the dynamics of single neurons and incorporating the results into his computer simulations, Sejnowski is hoping to develop a model and a theory for how spike-time training may be important to learning and sensory systems as well as understanding the role it plays in plasticity.
"What we've learned about the brain in the past ten years is more than we'd learned in all of history before then. We've only scratched the surface so far," he says. "I can't wait to see what the next decade brings."