The details of how proteins fold on themselves are shrouded in mystery. Scientists like David E. Shaw are using high-grade computers to cast light on those processes, but computing power proves to be a real barrier.
Shaw spoke at Friday’s Computing @ Exascale symposium at the Harvard School of Engineering and Applied Sciences where he showed attendees a simulation of what it looks like when a protein folds. The worm-like structure thrashed and twitched as segments locked into their final positions. In the end, the protein coiled into the shape that experiments have shown it should arrive at.
The video lasted perhaps 60 seconds, but creating that clip required a ludicrous amount of computing power. The individual frames, Shaw said, represented femtoseconds. That 10^-15 seconds.
Here’s how wikipedia puts that in perspective: “A femtosecond is to a second what a second is to about 31.7 million years.”
Stack on top of that the number of forces that the number of forces that have to be simulated for each atom in relation to each other atom, and you start to get a sense of how quickly trying this at home would brick your laptop.
Right now, his lab is working on a number of machines that can build simulations up to a millisecond in length. That’s one one-thousandth of a second. Many proteins take longer than that to fold, and shaw said he eventually wants to simulate processes that take much longer to complete.
“There are things we want to know that might take a million times or a billion times longer,” Shaw said. “Some things happen over the course of weeks.“
But he’ll have to wait for computers to catch up with him. Even the scale of computing that he and others are talking about at today’s conference is just 1,000 times faster than today’s fastest machines. To achieve the kinds of things he’s aiming at, he would might need something a million or a billion times faster.
- What Would You Do With an Exascale Computer? (gistpodcast.com)
- Harvard researchers fold proteins with D-Wave quantum computer (gizmag.com)
- GPU Technology Conference Agenda Chocked Full of Supercomputing Sessions (insidehpc.com)
- TACC supercomputers simulate organization of membrane proteins at cell surface (phys.org)
- NVIDIA, IIT collaborate to develop supercomputer (rediff.com)