By: Maeve O’Connor
Dr. Stephen Keeley, a professor of natural sciences at Lincoln Center, has spent the last five years researching how to be a problem solver. Keeley is a computational neuroscientist whose work deals with the intersection of mathematics, statistics, and neuroscience. His current research focuses on developing statistical techniques to understand neural data broadly. Keeley said he works to learn “how entire systems in the brain work to do the thing a healthy brain does.”
Keeley received his bachelor’s degrees in neuroscience and physics at the University of Rochester before completing his Ph.D. in computational neuroscience at the Center for Neural Science at New York University. From there, he was a postdoctoral researcher at the Princeton Neuroscience Institute and dealt mainly with the statistical side of computational neuroscience.
The research Keeley is currently conducting is a continuation of research he has been working on for the past five years. His research does not attempt to answer a certain hypothesis surrounding neural activity; rather, it works to “come up with general techniques to understand many systems in the brain.”
“My research is really developing a set of techniques to address a current problem in the field. The problem is there has been a lot of engineering accomplishments over the last 10 to 15 years in neuroscience,” Keeley said. In the 1970s and 1980s, neuroscientists were only able to record information from single neurons, the cells that make up the brain. Over the last few decades, neuroscientists’ ability to record neurons has exploded and research is now able to be done on numerous neurons at the same time.
“It is now the case that labs will record from up to 1000 neurons at the same time and there are a variety of new techniques that will record these neural populations,” Keeley stated. However, the information provided by studying these populations of neurons can be difficult to understand. Neurons release “bursts of energy” known as action potentials. When researchers are tasked with recording these spikes of energy, it is difficult to make sense of this neural activity with outdated mathematical procedures for recording and analyzing data. This is where Keeley’s research comes in; “I’m not necessarily developing a technique to answer a specific hypothesis in a specific experiment. I’m hoping to develop a technique a lot of neuroscientists can use, that can help them all address their hypotheses,” he stated.
The technique Keeley is researching is used to understand systems of the brain, such as the auditory and visual cortex, which is a field of neuroscience known as systems neuroscience. Keeley said, “I hope that the kind of work I do can be general, that it can be a technique you might want to use if you want to find out how the visual system works or the auditory system works.” Keeley continued, “In that way, the kind of work I’m doing is not driven by understanding a particular system, it is driven by trying to come up with general techniques to understand many systems in the brain.”
Keeley’s research is very collaborative. The mathematical techniques he has developed during his research are able to be implemented by experimental neuroscientists when they analyze their research findings. One of the models Keeley uses to develop mathematical techniques is called a latent variable model. This model works to analyze the action potentials of a large population of neurons and compress the entire population’s activity into smaller groups based on the similarities and differences among the neurons’ rates of activity. In the example Keeley used, the population of neurons was 1000. “You can imagine that if I record from 1000 neurons, 500 neurons have one rate of activity. … And then maybe there’s another rate of 500 neurons that does the opposite.” What Keeley’s mathematical technique will do is “take that 1000 neurons, find those two populations, and extract the firing rate of population one and population two.” With these new sets of populations, Keeley is able to go back to the experimental neuroscientist and say, “We have found structure in the data.” Keeley’s technique provides experimental neuroscientists with a better understanding of neural activity in the populations they are researching, which helps them understand broader information surrounding the brain’s systems and functions.
One of Keeley’s recent developments uncovered an interesting finding of the visual cortex. Say an individual is shown a two-minute video, a stimulus. They wait five minutes and are then shown the exact same video again. The individual recognizes they are watching the same video; however, their neural activity changes each time the stimulus is played. These results present an interesting question for neuroscientists: what about the activity is the same with each presentation, and what about the activity is different?” Keeley has developed an algorithm that extracts both populations, those that are different and the populations that are the same. Through his research, Keeley has discovered that the different populations do not interfere with one another, meaning the patterns that are the same and different across each stimulus are not affected by the deviations created when the stimulus is presented.
Keeley said it is helpful to explain his findings by visualizing the compressed neural activity, the action potentials, as spots covering a piece of paper. The spots cover the piece of paper during every presentation of the stimulus. However, there are deviations in the spots, or neural activity, that also occur each time the stimulus is presented. Keeley says, “During the deviation, what we can show is that the differences on the trials don’t change the location on the piece of paper.” “There’s a way the stimulus is represented in the neural activity that is not affected by the deviations on each trial,” he continues.
Dr. Keeley’s work is vital to better understanding the brain systems that shape an individual’s behavior. His collaborative research has surely prepared him for working with students as a professor in the Department of Natural Sciences. It will be exciting to see what he accomplishes in the classroom, the lab, and maybe even the auditory cortex here at Fordham.