In the past few protein folding competitions that Rosetta@home has participated in, it has emerged among the top performers.
I understand that predicting protein structure relies on starting from many random points and comparing the energy levels from these starting points to find the optimal shapes. Given the huge computing advantage of rosetta@home (due to the large numbers of volunteers), as compared to less popular projects, say, predictor@home, how is it that any conclusion about which method is the best can be drawn?
Seems to me that the competitions like CASP are not really fair.. What if better methods to predict folding are abandoned due to this?

---

In the past few protein folding competitions that Rosetta@home has participated in, it has emerged among the top performers.
I understand that predicting protein structure relies on starting from many random points and comparing the energy levels from these starting points to find the optimal shapes. Given the huge computing advantage of rosetta@home (due to the large numbers of volunteers), as compared to less popular projects, say, predictor@home, how is it that any conclusion about which method is the best can be drawn?
Seems to me that the competitions like CASP are not really fair.. What if better methods to predict folding are abandoned due to this?

Predictor@home could easily get more activity from participants if they were generating more workunits. The last time I looked, they were only recycling a few of their old workunits that had failed before, and only about one of them every third day. In other words, they've been essentially inactive for several months. I've been registered as a participant there for months, and haven't gotten even a single workunit from them yet.

That's a good question, Sam.

I'm pretty sure that internally the project is testing different versions and optimizations in "apples vs. apples" tests to see which ones give the best result from the same starting points and with the same amount of CPU power.

The real question is: Is CASP a competition between different labs/projects, or a competition between different techniques. If it's the former, than however much CPU power you can throw at the problem is fair game. What matters is results.

---

Actually, the real question is, is your approach good enough to cure or immunize against the subject disease. I mean if you find treatments for cancer or HIV, does it matter how much compute time went in to the solution? Since none of the methods used are reliably that good, events such as CASP help to keep tabs on what the state-of-the-art is, and to motivate further improvements.

Just keep in mind that similar outcomes of CASP were occuring well before Rosetta@home was introduced. This is part of why BakerLab has grown the way it has. Not the other way around. And if another idea has merit, you should be able to demonstrait it against some more limited set of proteins with a significantly limited set of available processing power.

---

Rosetta Moderator: Mod.Sense