Can Proteins From Living Cells Solve Problems That Vex Supercomputers?
When nature knows best
/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer/78/93/7893995a-9dc7-4b23-9330-db4481b61879/may2016_e06_phenom-wr.jpg)
Our daily life is so digitized that even technophobes know that a computer is a bunch of electronic transistors that process 1 and 0 signals encoded in a program. But a new kind of computing may force us to reboot our thinking: For the first time scientists have tapped the energy source used by living cells to power tiny proteins to solve a math problem.
The research, led by a father-son duo, is a boost for biocomputing, which promises devices that tackle complex tasks and use much less energy than electrical machines. “It’s not a question of making faster computers,” says Dan Nicolau Jr., lead author of the new study, who earned a PhD in mathematical biology at Oxford. “It’s a question of solving problems a computer can’t solve at all.”
Take code-breaking, which can involve sifting through trillions of combinations to reach one correct solution. Perhaps surprisingly, mainframe computers aren’t so great at solving a problem like that because they tend to work linearly, making calculations in one sequence at a time. Parallel processing—trying multiple possible solutions simultaneously—is a better bet.
Which is where the new experiment comes in. For years, Dan Nicolau Sr., head of bioengineering at McGill University in Montreal, has studied the movement of cytoskeletal proteins, which help give cells their structure. Around 2002, his son, then an undergraduate, was thinking about how rats in mazes and ants on the hunt solve problems. Could the proteins that his dad researched also be put to work solving puzzles?
To test the question, they first had to translate it into a form that the proteins could react to. So the researchers chose a mathematical problem, plotted it as a graph and then converted the graph into a kind of microscopic maze, which was etched onto a one-inch-square silica chip. “Then you let that network be explored by agents—the quicker, the smaller, the better—and see where they’re getting out,” Nicolau Sr. says. In this case, the agents were cytoskeletal protein filaments from rabbit muscle (and some grown in the lab), and they “explored” the various solutions of the maze, like a crowd looking for exits. Meanwhile, the meandering proteins picked up energy from the breakdown of ATP, the energy-releasing molecule that powers cells, and the “answers” emerged from watching where the proteins escaped, then retracing their steps.
This experimental biocomputer can’t outperform an electronic machine, and it’s designed to solve just one problem. But researchers think the concept can be scaled up someday to tackle challenges that currently befuddle conventional computers, using “thousands of times less power per calculation,” says Nicolau Jr. Cryptography, drug design and circuit paths all pose big mathematical challenges that are just begging for a natural parallel processor. And as Nicolau Jr. says, “Life does things more efficiently.”