Having experience in this field, I'm usually very cautious about the sort of over-exuberance displayed in this article. First of all, the title is misleading. The 100,000x speed up is vs the uncatalyzed reaction, not vs some other enzyme. As enzymes go, that makes the man-made protein still about 4-6 orders of magnitude (10,000-1,000,000x) slower than the best enzymes.

There is a lot of interesting research going on in this field, but I worry that, like much of modern science, the focus is too much on "how can I turn this into a product" and less on "have we answered all the questions". The way I was taught was that Scientists answer the questions, and Engineers come up with the applications. What we have today is a lot of Scientists acting like Engineers.

One such unanswered question: How do we factor in entropy? (I did my Ph.D. prelim exam on that topic... If anyone is interested -- http://github.com/jballanc/prelim/raw/9a4ab5287b6776b9987834...) What's interesting about that topic is the way that we're applying new approaches and algorithms to the questions being asked. So, while the future of this field is certainly bright, we're still at the head-down keep-pushing-forward stage...

...keep the champaign on ice for now.

The article also says that no known natural enzyme catalyzes this particular reaction, so it's still a nice achievement, even if the artificial enzyme is a poor performer.

I've been running Rosetta@home since 2005 (now on about 16 cores total), and I strongly encourage everybody here to give it a go.

Such a shame to waste idle CPU cycles... It might take a bit more electricity, but buy a few CFLs to compensate or something. This is important science.

I did my dissertation work in this field, and I wish I could properly communicate to everyone how not important this stuff is to science. There's no doubt that this is an interesting engineering problem, and that it has a gee-whiz aspect that has fascinated technology writers for decades. But most of these guys aren't learning anything from the work that they're doing. It isn't science.

There are plenty of researchers learning new things with computers -- computers are used to assemble DNA sequence reads, find genes, analyze mutation hotspots, on and on. But protein design -- even when successful -- doesn't teach us anything new. It's a random search through the dark with a tiny flashlight, and success primarily means waving the flashlight around faster. There are very few take-home lessons from a successful design, other than "that seemed to work". That's why these researchers are so focused on hardware, and sampling more structures in less time -- it's all about the million monkeys.

This may sound harsh, but I sincerely believe that the electricity that you save by turning off your PC when idle would be far more valuable to the world than any marginal technological gain you're giving these researchers. Please don't fall for the hype that you're curing cancer or HIV with this stuff.

Thanks for saying what you just did. Although my protein structure prediction days are far beyond me, cycles for projects like these are, for the most part, not going to result in major scientific advancement. In most cases, they are a proof of concept, but you really can't do any production work using spare home computers. Now research on accelerated hardware, new scale out programming models, better algos is all worth it.

I will disagree on the utility of figuring out a structure though and this is based on experience where there was always a dual component to research, which the prediction guiding, for example, site-directed mutagenesis experiments. The computational design is a lot cheaper and can really help guide project direction.

And in the long run, there is nothing wrong with just understanding the physics of folding. The problem is we really aren't there yet, but that doesn't mean we should stop