To keep the original thread cleaner, I've locked it, and opened this thread for discussion and questions about the CASP8 stories.

--Mod.Sense

---

Rosetta Informational Moderator: Mod.Zilla

tell us everything! ;)

---

The more news the better!
We are starved for information about what has been learned or ideas to applied in the future.

Also, what do people want to see? Do you like seeing superpositions of successful predictions? Do you want to hear more about the open problems in structure prediction and what we're pursuing? We also have some other anecdotes and stories from CASP8 if people want to hear that as well.

I think many people have trouble seeing the superpositions, especially when they are very close. If you could allow us to see a superposition within one of the 3D modelers that might help. Otherwise, I think sometimes graphs are more easily understood. Graph of RMSD of Rosetta's various models, or Rosetta's submitted models are compare to the other CASP submissions.

Yes, if you pinpoint the trouble spots of structure prediction, it will start a flow of ideas. And questions that help us all to better understand the challenge.

I for one, would love to hear how the process works within your lab. From the time a target is released, what steps do you take to solve the protein? How do you decide which of your many techniques to apply? How are initial results fed into next steps? How do you decide if you should attempt modelling with constraints? How much human experience goes into planning the work before it goes out to Rosetta? How many different approaches are applied to the work that is sent out? Do different people in the lab have different nitches? Such as proteins with less beta sheets? or proteins with symmetry? How do you know, in advance, if a given protein will have symmetry?

I think many of us have played Fold.it!. Perhaps you could offer some references to what people can actually see in the game. Is the comparative modeling what "rebuild" is doing in the game? If the structure is unknown, how do you come up with the constraints? And how do you know (or at least gain some confidence that) your chosen constraints are not omitting the native structure?

Please, as you post the explainations of new terms, bold them. Such as comparative modeling. That will make them easy to pick out when people review the thread in the future.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

The CASP 8 predictions from all participating servers are available at http://predictioncenter.org/download_area/CASP8/server_predictions/. If you're using Windows, the *.tar.gz files there can be unzipped and untarred (i.e., accessed) by using a program like 7-Zip; Mac and Linux users have built in utilities to deal with .tar.gz files.

And while you would need to register for an account (also publicly available), the predictions are also more conveniently available at the CASP site by navigating on the left menu to CASP Experiments -> CASP8 (2008) -> Predictions -> Server Predictions.

At that page, there are four servers clearly belonging to the Baker group: BAKER-DP_HYBRID, BAKER-GINZU, BAKER-ROBETTA and BAKER-ROSETTADOM. Which of these used Rosetta@home resources? It would also be helpful to know those servers' ID numbers, which I imagine are needed to figure out who made what predictions in the files I linked to.

---

David has suggested that several other members of our group post in this thread as well to talk about what we did, how we did it, and tell you some hopefully amusing anecdotes about our work during CASP8. Expect those as well in the coming weeks, along with more discussions of protocols, results, and future goals for the project.

Great - there hasn't really been much good information coming out of the project lately (I know you are all busy)... So this is a great idea. I'd love to hear anything about how CASP8 worked, etc, especially anything specifically to do with our work on Rosetta@Home.

---

Alver Valley Software Ltd - Contributing ALL our spare computing power to BOINC, 24x365.

I'd definitely like to know more about CASP8 in general, especially about the implications of further improving predictions (how far are we from reliably designing drugs?).

Would also love more updates in general, even if they are short. It really helps keep morale up, IMO. Nothing worse than seeing that the "news" section or Dr. Baker's journal hasn't been updated in over a month.

---

Let's hear more! Whatever you want to talk about! Still, if you're stuck for a topic, I'd be most interested in hearing about the limiting factor affecting the accuracy of protein structure prediction: would it be the Rosetta model itself or the computing power available?

---

I'd be most interested in hearing about the limiting factor affecting the accuracy of protein structure prediction: would it be the Rosetta model itself or the computing power available?

Both are limiting factors. With more computing power, the number of decoys increases (thus the chance of finding a very close-to-native structure increases). But computing power isn't everything. There are groups in CASP with fewer average FLOPS that do better than Rosetta depending on the category of prediction (e.g., either comparative or de novo prediction). So Rosetta's methods could always be improved.

On a separate note, I've got a few questions about Foldit. Could there be more explanation on how it was it incorporated into CASP8? Are its predictions available for viewing? Is it one of the four servers listed in my previous post?

And finally, for anyone interested there are some preliminary CASP8 assessments available here: http://www.reading.ac.uk/bioinf/CASP8/

---

Some nicely done graphs and superpositions from POEM's efforts are currently in their results page.

Did well with fragment fitting but could clearly be better when it come to de novo prediction. They have substantially less firepower then Rosetta of course.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

I wonder how the quality of results scales with crunching power.

With everything else being the same, does a doubling of power gives on average 2x 'better' results (not sure what metrics to use there), or is there a diminishing returns thing where 2x more power gives you 10% better results and so on?

---

I wonder how the quality of results scales with crunching power.

With everything else being the same, does a doubling of power gives on average 2x 'better' results (not sure what metrics to use there), or is there a diminishing returns thing where 2x more power gives you 10% better results and so on?

This is a bit speculative, but often in problems of this sort the accuracy (using the word a bit loosely) increases as the square root of the sample size. So R@h would have to send out 4 times the number of decoys to get a model that was twice as good, assuming this relationship holds.

---

So R@h would have to send out 4 times the number of decoys to get a model that was twice as good, assuming this relationship holds.

Hmmm, POEM 9.556 TFlops, Rosetta 65.310 TFlops.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

More CASP8 stories please!

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

We have to be careful about the comparisons being made with POEM targets at this point. They did better than the mean, but we don't exactly know the other competitors and how they compete. More importantly, those results only offer some insight on TWO targets. We also don't know much about the dificulty level of the targets poem posted to crunch.

With that said, I would like to see a comparison (Rosetta vs. POEM) once the results are reported. Have been busy at work so have not had the time to read the stories. Will hopefully have a follow-up soon.

---