I liked a project comparison/summarization writen by alex@simap so well that I swiped it as a start on the answer to this question in theBOINC Wiki. I am sure in time we will be adding to this, but, I like to capture good things like this for posterity.

Format has been changed slightly, more links added, spelling checked (of course, I may have changed correct spelling to wrong).

See: "What is the difference ..." for the article as it stands now.

I will be posting this in a couple other places to elicit comment. If there is more to say, or something should be changed, p.d.buck@comcast.net works ...

So, project guys, if you want something else stressed, more added, some removed ... nows your chance ...

---

That is a good start at describing the scientific differences between the projects.

I'll describe some other differences between the only two of the protein projects that I know myself. I think these thiongs should be documented as well as the science, though if they belong in the same part of the wiki is another issue.

There are also human differences - at present Rosetta seems to have a much better scientist to participant interface than Predictor, though I remember days when P was almost as good as R is now.

There are also practical differences for crunchers. Both P and R have some strange bugs and it may be that it is not wise to run them both on the same box. P tends to have bugs that make it less good at sharing a machine with other projects than R.

P checkpoints every 5 - 7 min on a 2.8GHz box, R currently (Jan 2006) waits about half an hour usually, but can wait over an hour before checkpointing (both checked using Linux on a 2.8GHz box with HT enabled, and using with the command line interface track checkpointing). This can make a big difference if the box is switched off frequently, or rebooted between Windows and Linux at frequent intervals during the day (a practical issue for some of the boxes running BOINC for me). It matters (a little) even if the box keeps office hours.

River~~

---

I believe I now have a rough idea how the Rosetta algorithm attempts to fold proteins but I am still completely in the dark about the approach used by Predictor. Unfortunately the Predictor home page doesn't seem to provide any substantial information about the algorithm used (and of course the text that is now in the Wiki doesn't go beyond what is avaliable on the Predictor website). It would be nice to see a comparison between the Rosetta and Predictor methods highlighting the merits and shortcomings of each. I guess it would make more sense to post this over at Predictor (I don't have an account there) since it mainly seems to be info on the Predictor algorithm that is lacking...

Sorry about the double post.

I believe I now have a rough idea how the Rosetta algorithm attempts to fold proteins but I am still completely in the dark about the approach used by Predictor. Unfortunately the Predictor home page doesn't seem to provide any substantial information about the algorithm used (and of course the text that is now in the Wiki doesn't go beyond what is avaliable on the Predictor website). It would be nice to see a comparison between the Rosetta and Predictor methods highlighting the merits and shortcomings of each. I guess it would make more sense to post this over at Predictor (I don't have an account there) since it mainly seems to be info on the Predictor algorithm that is lacking...

I can't help much here as I don't know what methods Predictor is testing either.

---

I can't help much here as I don't know what methods Predictor is testing either.

Based on a review of their site, I am not sure they know either ... :(

---

I can't help much here as I don't know what methods Predictor is testing either.

Based on a review of their site, I am not sure they know either ... :(

Thanks Paul. This one post just about made my evening. I was down in the dumps, but this made me laugh. Never lose your sense of humor.

Kathryn

---

Kathryn :o)
The BOINC FAQ Service
The Unofficial BOINC Wiki
The Trac System
More BOINC information than you can shake a stick of RAM at.

I believe I now have a rough idea how the Rosetta [snip] info on the Predictor algorithm that is lacking...

Ive put a posting here on the P@h site, quoting the whole of your posting Hoelderlin. Let's see how they respond over there

River~~

---

I can't help much here as I don't know what methods Predictor is testing either.

Based on a review of their site, I am not sure they know either ... :(

Thanks Paul. This one post just about made my evening. I was down in the dumps, but this made me laugh. Never lose your sense of humor.

Its about the only thing I have left ...

We will see if River~~'s post will wake up anyone at PPAH ... it has not been the same since MT left (can't remember how to spell her name). Ah,well ...

Anyway, I am glad you got a lift out of it ... me, made me more depressed (does not seem possible somehow).

---

Another good comparison between the protein projects is here - including responses from the scientists at both the Human Genome project and Bakerlab about their different approaches to the Rosetta program.

Don't know if you've seen this before, Paul? I hadn't. Thanks to Dimitris Hatzopoulos of Einstein@home for this link :)

Still no response from the Predictor scientists by the way - I have just bumped the thread over on their boards but will not be bothering to remind them again...

R~~

---

don't forget, there's also the United Devices project

In the past there's been smallpox, anthrax and oncology therapeutic agent crunching done. Presently, they're engaged also in Rosetta proteome folding. However, its done entirely different than either P@H or R@H. World Community Grid BOINC client is also doing something similar to what UD's client is processing. However, its different too. And as far as I understand, WCG will soon be doing Aids@Home.

---

and we now have a reply from Predictor, here

R~~

and we now have a reply from Predictor, here

R~~

...great, this is indeed the most detailed comparison between the two projects I have seen so far and thanks again to River~~ for cross-positing my original question. :-) Lets see what Prof. Brooks will have to say about the detailed CASP6 comparison...

-H.

and we now have a reply from Predictor, here

R~~

...one thing that wasn't addressed in the response by Chahm are the different computational requirements of the two methods. It seems to me that the Predictor method would be computationally more demanding (higher dimensionality of lattice space in Predictor as compared to "fragment space" in Rosetta, and molecular dynamics is more demanding than minimizing the energy in Rosetta's all atom relax phase).

So, project guys, if you want something else stressed, more added, some removed ... nows your chance ...

...and also thanks to Paul D. Buck who started this discussion by creating this thread (and by including a detailed project comparison in the Wiki). Lets see whether we can get him to rejoin these boards ... :-)