In one form or another, Big Data has found a role in academic disciplines from literature to particle physics. New computing machinery, blazing fast and muscular, puts access to unimaginable quantities of information at the tip of a mouse.
Paul Valiant is intrigued by the sudden emergence of the Big Data metaphor and how it intersects and informs his research. “The thing that surprised everybody about Big Data was that for some problems all you have to do is use known methods — but apply them to the entire content of Wikipedia and half a million books scanned. A computer can win Jeopardy,” he said, referring to IBM’s Watson computer system that faced off against Jeopardy champions last year and won.
“For some problems, the data does stuff for you that people used to think required a conceptual breakthrough.” Another example he raised is the recent data-driven success of tools to translate text between English and other languages.
Whether mediated by Big Data or not, the spread of computation — computer science — across the academic landscape and deep into disciplines presents opportunities for conceptual advances. “As the long synergistic history of mathematics and physics attests, the benefits flow both ways,” he once wrote. “Newton’s mathematical accomplishments, for example, cannot be understood except in light of the developments in mechanics he was pursuing. Thus it may be hoped that when the interconnections between computation and the other sciences have reached a similar level of maturity, corresponding fruits will have been gained on both sides.” The essential philosophy, he said, is to engage, to be willing “to get one’s hands dirty in the service of the right problem.”
Protein folding is a case in point. Proteins are built as long chains of amino acids which interact with each other to fold up into an arsenal of molecular machines which collectively run our bodies. Correct folding is essential, akin to the correct assembly of an industrial machine, but the body has no blueprints; its machines must assemble themselves following trillions of exquisitely sensitive interactions of the laws of physics. There has long been the hope that computers could predict how proteins fold, sidestepping a painstaking protein-by-protein effort of biology labs worldwide, but current approaches would still take hundreds, even thousands of years to simulate the single second that a typical protein takes to fold.
Conceptually, however, there is strong evidence of exploitable structure — natural rules that would substantially reduce the complexity of the search for the folded state. The Foldit Project, a videogame-like crowd-sourcing Internet site, asks users to fold components of a protein in ways that require the least amount of energy — and then scores their attempts.
“I'm not a user-interfaces guy or a crowd-sourcing guy. I'm a theory and algorithms guy,” Valiant said. “But what Foldit demonstrates is that there is structure to the problem. We were’t sure if there was structure to grab onto, but the fact that humans can do it at all is proof of concept. Foldit tells us that there are algorithmic breakthroughs waiting to happen.”
Toward that end, Valiant will be offering a graduate-level course in protein folding this semester, but in an advanced, real-time, 3-D way. “Seeing things in 3-D helps your brain, so we'll give the brain the tools it wants,” he said. “I've bought some 3-D glasses, those red and blue things, which I'll hand out to students. They’ll have six-degrees-of-freedom joysticks to push, bend, or twist the protein in three dimensions. And we’ll do it live at body temperature.”
Valiant’s path to an assistant professor appointment at Brown began early, growing up in a household fascinated by math and science. There was a second-grade project in the finer points of Microsoft Basic. In his early teens, there was an encounter with a college-level computer science textbook by Andries van Dam and John Hughes, a gift from his father. At Stanford, there was undergraduate work in physics and math (B.S., physics and math, 2004), but a graduate-level turn toward computer science (M.S., computer science, 2004). He completed his formal studies at MIT (M.S., computer science, 2007; Ph.D, 2008) and did postdoctoral work at MIT and the University of California–Berkeley, his last post prior to Brown.
What brought him to Brown? “It had something to do with the department being small and communicating very well,” he said. “You get the sense that professors around here are very willing and able to trust each other’s judgments, to work with each other on projects. Things seem to work very fluidly.”
That and a sense expressed by former department chair Eli Upfal that Brown’s collaborative culture would be a wonderful fit for a young computer scientist deeply involved in “interdisciplinary theory.”