I first started thinking about proteins while confined to tiny tent on a rainy afternoon in the Rockies. On the previous day one of our group had to be helicoptered out at 13,000 feet, when he fell and broke something. After all that excitement, a nice day hike would have been nice. But the weather didn't cooperate and, not having packed a book, I started thinking about the folding problem.

There are actually two protein folding problems, and before I even get into that, I should first explain the limited scope of what I'm calling the folding problem. Protein molecules can perform surprisingly complex tasks, and this requires structures that often are complexes comprising multiple molecules and that can have multiple folded shapes or conformations. Even Nature has trouble folding these high-end machines, and has evolved special chaperone proteins to help in that case.

I'm interested in a smaller, simpler class of molecules called globular proteins. Their simplicity in no way diminishes my respect for these marvels of engineering. Imagine for the moment you are a simple organism: a struggling start-up with an enormous product line of widgets and not much in the way of skilled labor. In fact, to stay competitive you have had to down-size to just a single type of robot, the RIBOSOME, capable only of stringing together chains of raw materials that it selects from some 20 bins. But miraculously, or thanks to the laws of physics, these chains spontaneously fold up into unique, functional widgets with essentially 100% yield!

Simple globular proteins (SGPs) are a Google-X approach to manufacturing. Products are fabricated from linear strings of amino acids and self-assemble into unique, 3D structures when they leave the factory. This brings me to the first of the two protein folding problems: the rules. Self-assembly happens on a reasonable time scale in vivo because Nature's clock rate, as implemented by electro-chemical rules, runs in the terahertz range. This has not stopped ventures such as folding@home, who by recruiting all the idle computers on the planet attempt to fold proteins by following Nature's rules. And folding by accurately mimicking Nature is also what you want to do when studying how folding goes awry, as has been proposed as the mechanism for certain neurodegenerative diseases. But for non-pathological folding of SGPs I'm betting there must be some slicker rules that work just as well.

Once you've settled on the rules, the second protein folding problem is finding a fold that satisfies those rules. The algorithm Nature uses, and that applies to its electro-chemical rules, is simply to step forward in time the protein molecule and surrounding solvent by the laws of mechanics. When properly folded, a protein has a lower free energy than when it was ejected from the ribosome, and SGPs, it is generally believed, have a unique, lowest free energy folded form. My proposal, for a better folding algorithm, is to exploit this special property. Or that was my vision, as I was watching those rivulets meandering down my tent on that rainy afternoon many years ago. The equilibrium state of a folded protein isn't the usual sloppy compromise between energy and entropy. Rather, it is a cleverly designed jigsaw-puzzle-contraption where each part is neatly snapped in place. 

And there you have it: a crazy research project which proposes to solve complex protein folding "puzzles" according to rules that have to be worked out as well!