Liam Paninski: Neural Code Breaker

© 2007 Greg Kessler Photography

The lone distraction in Liam Paninski's austere New York City office is an acoustic guitar lying strings up on his desk. A computer idles beside the instrument, and empty bookshelves line the room that the soft-spoken Columbia University professor of theoretical neuroscience jokingly calls his "lab."

Paninski's decor parallels his approach to neuroscience. The statistical models and computational analyses he's been developing are meant to clarify the voluminous clutter of data pouring in from increasingly sophisticated recording equipment. He's been working to streamline the cacophony.

"This is really becoming a huge problem in neuroscience," says Paninski. "There are lots of [labs] where people have collected gobs of data, and we really need to develop tools that can handle that." Developing these tools has been Paninski's single-minded focus, fostered since he was an undergraduate in John Donoghue's lab at Brown University.

Nicholas Hatsopoulos, a postdoc in Donoghue's lab and Paninski's mentor, was immediately struck with the undergrad's mathematical acuity. "It was amazing," Hatsopoulos says. "I remember trying to teach [Paninski] about information theory, and he didn't know anything. Within two weeks, he knew more than I did."

Paninski published much of his undergraduate lab work, even coauthoring a Nature paper that described a mathematical approach used to decode macaque brain waves. With it, the monkeys could control computer cursors deftly, using only their brains. ¹

At Brown, Paninski also developed a novel experimental design that "allowed us to use the right mathematical tools in the right way to study brain coding," says Donoghue. ² Today, neuroscientists still use this design.

Through his PhD at New York University and subsequent postdocs both there and at University College London's Gatsby Computational Neuroscience Unit, Paninski continued to develop the robust statistical tools needed to clear the clutter of neuroscience data. As a grad student, he turned his attentions to the study of how neurons in the retinal ganglia organize and send visual information to the processing centers of the brain. ³ "To a statistician, this is beautiful data," Paninski says. "You have the ability, the potential, to understand all the information the eye is sending to the brain."

The focus and drive that has characterized Paninski's career thus far has not subsided in his latest role. Sean Escola, a student in Columbia University's MD/PhD program, works often with Paninski and tells of their recent trip to this year's Computational and Systems Neuroscience Conference. Veteran researchers attending the conference typically showed one poster, says Escola. Paninski had 13.

This doesn't surprise Paninski's former colleagues. "I knew he was on a course to becoming a real star," remembers Donoghue of his previous undergraduate student. "I expect it of him."