Are there any objections?

If I hear none prior to 3 hours from the date stamp on this message, I will proceed.

A fine idea. Takes too long to display and reply to this thread now. Perhaps do the same with Dr. Baker's thread? Or even color code his posts interspersed within the same "archieved", read-only thread??

---

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/

This discussion is continued in This Thread.

Please make all new posts here.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

The 5.16 release log says "...new feature where at the end of a simulation, Rosetta compares its fold to the predictions made by a dozen other algorithms." I can see that if several methods of prediction agree, it tends to strengthen the validity of the results. How are the results of all the other methods already known? Is this a CASP thing to try and get the best model? Is Rosetta making adjustments to bias some of it's decisions based on the other results (i.e. changing it's prediction slightly to yield something that more closely matches)? Or is it just comparing and scoring the fit with the other predictions and returning the scores?

---

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/

I had an e-mail from CASP this morning that said the results of the robots would be made available after something like 3 days of issuing the protein. Perhaps these results are being collated and bundled with the wu's? Should be straight forward toi do, the result files are very small.

Of course, I could be entirely wrong.

I am always wary of using homologs and other predictions. I can see cases of, "we all made the same mistake", or, "we searched the wrong domain because the other models pointed to the wrong place", this kind of thing.

---

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

In response to some of the recent discussions on the science boards, today I'd like to tell you about how Rosetta is being used to help understand diseases caused by protein misfolding.

A significant fraction of human diseases are caused by proteins misfolding to form long "amyloid fibrils". These diseases range from Alzheimer's disease to infectious diseases from amyloid forming prion proteins. A huge breakthrough in the understanding of the process of protein misfolding to form amyloid fibrils was published in Nature last year from David Eisenberg's research group at UCLA. They reported the first high resolution structure of an amyloid forming peptide. It revealed a set of interactions which seem very likely to be general to most if not all amyloid structures.

We have been collaborating with Eisenberg's group to try to predict the portions of proteins known to form amyloid structures responsible for amyloid fiber formation. We use the rosetta-design method to identify sequences compatible with a generalized model of their amyloid structure. You can read about the promising results of this work in the collaborative paper with Eisenberg's group that is posted on the "2006" portion of our home page publication list mentioned in my previous post. The next challenge which we are collaborating on is to design "caps" that will add on to fibers and prevent them from growing further. This is a good example of how basic research development can have applications to pressing medical problems that were entirely unanticipated.

On a different note, Rhiju asked me to request that users with computers with remaining problems sign up for ralph@home--the error rate is significantly lower on ralph perhaps because there are a larger fraction of high end machines, and this makes it harder to track down the remaining issues on rosetta@home.

Dear Dr Baker,
One quick question. I am still wondering if Rosetta is working parallel to protein misfolding cyclic amplification techniques!
thanks

---

Refering to my comment above, here is the e-mail I received from CASP.

Next week we are planning to release 8-10 targets again (based
on their availability and difficulty).

We want to remind all the participants that server predictions
are being made available 3 days after the target is released.
They can be accessed from the main casp7 web page by following
the link "Server Models". Note, that tarballs of all 3D models
(with AL predictions already converted into coordinates) are
posted at the same place usually with a short delay (1/2-1 day).

I'd have included it above, but at the time, I was re-installing Windows on the machine that has my e-mail client on it!!!

---

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

Refering to my comment above, here is the e-mail I received from CASP.

Next week we are planning to release 8-10 targets again (based
on their availability and difficulty).

We want to remind all the participants that server predictions
are being made available 3 days after the target is released.
They can be accessed from the main casp7 web page by following
the link "Server Models". Note, that tarballs of all 3D models
(with AL predictions already converted into coordinates) are
posted at the same place usually with a short delay (1/2-1 day).

I'd have included it above, but at the time, I was re-installing Windows on the machine that has my e-mail client on it!!!

if all the automated methods such as robetta agree on the rough outlines, we do make use of this information--this is the "comparative modeling problem". Bin has developed some neat new approaches to this problem that I described last month (basically, the searching problem is similar to what you have been seeing all along, except that instead of starting with a fully extended chain we start with a variation on the starting model) that he is testing out on these cases (the challenge is to go from the rough and usually inaccurate starting model to an accurate high resolution structure). Rhiju is focusing on the case where the automated methods all disagree (like T283, 285 and 287) and for these the automated methods do not provide any useful information.

---

On the AP article:

Donna from the AP sent me a draft of her article; it is great! Thanks to all who helped her with it! She says the AP reaches half a billion people each day--hard to beat that for publicity!

Just a reminder to create a "landing page" specifically designed for this (a la BBC-CPDN's 1-2-3 join page and ideally an installer of BOINC+Rosetta), i.e. instructions for "regular", non-computer-savvy people.

PS: The LiveScience article -Rosetta@home in the news (republished on YahooNews!, FoxNews, MSNBC) must have been read by many people, but had minimal impact on TeraFLOPS (why, your guess is as good as mine).

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity

On the AP article:

Donna from the AP sent me a draft of her article; it is great! Thanks to all who helped her with it! She says the AP reaches half a billion people each day--hard to beat that for publicity!

Just a reminder to create a "landing page" specifically designed for this (a la BBC-CPDN's 1-2-3 join page and ideally an installer of BOINC+Rosetta), i.e. instructions for "regular", non-computer-savvy people.

PS: The LiveScience article -Rosetta@home in the news (republished on YahooNews!, FoxNews, MSNBC) must have been read by many people, but had minimal impact on TeraFLOPS (why, your guess is as good as mine).

we don't have a special landing page--does the volunteer team have ideas for this?

---

On the AP article:

Donna from the AP sent me a draft of her article; it is great! Thanks to all who helped her with it! She says the AP reaches half a billion people each day--hard to beat that for publicity!

Just a reminder to create a "landing page" specifically designed for this (a la BBC-CPDN's 1-2-3 join page and ideally an installer of BOINC+Rosetta), i.e. instructions for "regular", non-computer-savvy people.

PS: The LiveScience article -Rosetta@home in the news (republished on YahooNews!, FoxNews, MSNBC) must have been read by many people, but had minimal impact on TeraFLOPS (why, your guess is as good as mine).

we don't have a special landing page--does the volunteer team have ideas for this?

They were working on it, but the person doing the coding went on travel for a few weeks. He is back now. Perhaps they will have something in time.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

Many of the targets are very closely related to proteins of already known structure; in fact I'm not sure why the experimentalists bothered to determine their structures! The search is pretty easy in this case, and we are not putting too much effort into these predictioins (they are not so exciting).

...perhaps for some of the other teams these are to help them gauge their progress since the last CASP.

Hope to see some T296's on Ralph soon.

For those that were wondering, here is a thread from the AP reporter.

---

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/

One of the Rosetta team mentioned running 10k decoys, and using that data to create another WU to run 10k more decoys; in the hopes that the approach would result in better results than our running 100k decoys on a single WU. Are we going to try this approach on the 445 AA behemoth that you announced?
Have we got to try out the 10k, analyze, create a new WU and create 10k more decoy approach yet? (How well did it do?)

---

One of the Rosetta team mentioned running 10k decoys, and using that data to create another WU to run 10k more decoys; in the hopes that the approach would result in better results than our running 100k decoys on a single WU. Are we going to try this approach on the 445 AA behemoth that you announced?
Have we got to try out the 10k, analyze, create a new WU and create 10k more decoy approach yet? (How well did it do?)

We want to test this approach more systematically after CASP is over. Rhiju and I were talking today about what to do about the 445 residue monster; there are some hints it may be composed of two separate domains which have more tractable sizes; you will see work units from this divide and conquer approach soon.

---

On the homepage:

We have changed the workunit buffer size from 65,000 workunits to 20,000 workunits. This is so that CASP7 targets experience a smaller lag time between being queued and being sent out.

What is the impact for the users?
More risk that the project becomes without WU's?
Or...?

---

Member of Dutch Power Cows

What is the impact for the users?
More risk that the project becomes without WU's?

There is basically no impact to users. But people had noticed the drop and expressed concern.

If the work generator should happen to fail (haven't seen that happen yet), then yes, the work would run out sooner. But the way Rosetta works, we're all basically crunching on the same small number of proteins at the same time. We're just exploring a different area of them. So the WU generator can easily keep up. It just picks a different random number and sends another copy of the protein(s) du jour.

With the larger WU buffer on the server, they would add the latest CASP proteins to the list to crunch on, and not even send any OUT for half a day or so. With the smaller buffer, they'll be sending the new CASP targets sooner and getting back initial results sooner. This will help them determine if they want to try one of the other approaches in their toolbox to try and solve it.

---

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/

...This will help them determine if they want to try one of the other approaches in their toolbox to try and solve it.

Just as an aside, with the CASP models if I happen to see one of the workunits in my queue has completed, I manually "Update" BOINC so that it is sent in right away. I get the sense that the project team is interested in getting results back as soon as possible after a unit is released to us crunchers.

---

Regards,
Bob P.

I wrote an internal benchmark for Rosetta last week, and Rom now has a version that uses this to compute credits. Rom suggests however that we wait until after CASP to deploy it because it may take a few iterations to make it acceptable to everybody. I don't know how difficult it will be to "get it right", but I'd like to start testing it on Ralph soon.

I for one am in agreement with Rom, and would highly recommend not adressing this issue until your CASP predictions are in, I would also recommend that you don't release it on Ralph until CASP finishes as you may need to test a point release before CASP finishes (I'm not sure I made sense here, you don't want to clog Ralph with the new credit system as you may need to improve your main R@H algorithm during CASP)

The new version soon to appear on ralph will also have a fix Rom put in for graphics problems; as reported on the boards, a good fraction of the errors seem to be associated with the graphics (I suspect the fact that they consume lots of memory is part of the problem), and in the new versions graphics related errors should abort the graphics but not disrupt completion of the Rosetta calculation.

I have only seen ATI graphics problems to date (I don't pretend that I have seen every graphical error, but from what I've seen reported), I personaly think this is a BOINC/ATI problem.

I have 1GB of RAM on my main machine it has an ATI 9800 pro AGP 128MB videocard and while looking at the Rosetta@Home graphics a WU died. As I've said I have never seen any other graphics card have this problem.

Oh before I forget DR D.B. , would you write up a comparison between R@H and Bluegene when you get you hands on it, I would really like to know the differences in speed and such.

---

I wrote an internal benchmark for Rosetta last week, and Rom now has a version that uses this to compute credits. Rom suggests however that we wait until after CASP to deploy it because it may take a few iterations to make it acceptable to everybody. I don't know how difficult it will be to "get it right", but I'd like to start testing it on Ralph soon.

The new version soon to appear on ralph will also have a fix Rom put in for graphics problems; as reported on the boards, a good fraction of the errors seem to be associated with the graphics (I suspect the fact that they consume lots of memory is part of the problem), and in the new versions graphics related errors should abort the graphics but not disrupt completion of the Rosetta calculation.

How about instead of a "Internal benchmak" really counting flops
and using the proper boinc interface for this case ?
http://boinc.berkeley.edu/api.php

Credit reporting 
By default, the claimed credit of a result is based on the product of its total CPU time and the benchmark values obtained by the core client. This can produce results that are too low if the application uses processor-specific optimizations not present in the core client, is compiled with different compiler settings, or uses a GPU or other non-CPU computing resource. To handle such cases, the following functions can be used. 

void boinc_ops_per_cpu_second(double floating_point_ops, double integer_ops); 

This reports the results of an application-specific benchmark, expressed as number of floating-point and integer operations per CPU second. 
void boinc_ops_cumulative(double floating_point_ops, double integer_ops); 

This reports the total number of floating-point and integer operations since the start of the result. It must be called just before boinc_finish(), and optionally at intermediate points. 

Why Rosetta, currently, does not use any optimization ?

Using 3Dnow! (for Atlhon XP+) and (sse2 for Pentium IV & others)
can shirink the CPU time required to finish a float-point WU by 6 times (1:6)
So, why not ?

BTW: Was that "Internal benchmark" compared with a credited
benchmark program ?
eg: Sisoft Sandra
http://downloads.guru3d.com/download.php?det=177

ps: boinc is well know to produce very low (fantasy) benchmarks.

Do u believe,
that some team leader, wanting credits to place his team on top position,
will crunch by standard boinc benchmarks ?

*using a optimized boinc togheter with an optimized application my pc
can produce 22 credits/hour

*and I get this granted even on projects that do use "Quorum" and not everyone
is using "optimized" boinc/application, cause I claim the *same* credits
that someome else claim, using "standard" boinc/application

*Only they claim that *same* credits after 6 hours crunching a WU
while my pc is able to claim that *same* credits, at each hour -:)

Thus, I am not cheating in any way,
*and the importance of the "optimizations" !
that in the end, only causes science go a lot faster. and the CPU Hotter!


Thanks,

---

Click signature for global team stats

I wrote an internal benchmark for Rosetta last week, and Rom now has a version that uses this to compute credits. Rom suggests however that we wait until after CASP to deploy it because it may take a few iterations to make it acceptable to everybody. I don't know how difficult it will be to "get it right", but I'd like to start testing it on Ralph soon.

The new version soon to appear on ralph will also have a fix Rom put in for graphics problems; as reported on the boards, a good fraction of the errors seem to be associated with the graphics (I suspect the fact that they consume lots of memory is part of the problem), and in the new versions graphics related errors should abort the graphics but not disrupt completion of the Rosetta calculation.

How about instead of a "Internal benchmak" really counting flops
and using the proper boinc interface for this case ?
http://boinc.berkeley.edu/api.php

Credit reporting 
By default, the claimed credit of a result is based on the product of its total CPU time and the benchmark values obtained by the core client. This can produce results that are too low if the application uses processor-specific optimizations not present in the core client, is compiled with different compiler settings, or uses a GPU or other non-CPU computing resource. To handle such cases, the following functions can be used. 

void boinc_ops_per_cpu_second(double floating_point_ops, double integer_ops); 

This reports the results of an application-specific benchmark, expressed as number of floating-point and integer operations per CPU second. 
void boinc_ops_cumulative(double floating_point_ops, double integer_ops); 

This reports the total number of floating-point and integer operations since the start of the result. It must be called just before boinc_finish(), and optionally at intermediate points. 

Why Rosetta, currently, does not use any optimization ?

Using 3Dnow! (for Atlhon XP+) and (sse2 for Pentium IV & others)
can shirink the CPU time required to finish a float-point WU by 6 times (1:6)
So, why not ?

BTW: Was that "Internal benchmark" compared with a credited
benchmark program ?
eg: Sisoft Sandra
http://downloads.guru3d.com/download.php?det=177

ps: boinc is well know to produce very low (fantasy) benchmarks.

Do u believe,
that some team leader, wanting credits to place his team on top position,
will crunch by standard boinc benchmarks ?

*using a optimized boinc togheter with an optimized application my pc
can produce 22 credits/hour

*and I get this granted even on projects that do use "Quorum" and not everyone
is using "optimized" boinc/application, cause I claim the *same* credits
that someome else claim, using "standard" boinc/application

*Only they claim that *same* credits after 6 hours crunching a WU
while my pc is able to claim that *same* credits, at each hour -:)

Thus, I am not cheating in any way,
*and the importance of the "optimizations" !
that in the end, only causes science go a lot faster. and the CPU Hotter!


Thanks,

I beleive Rom knows what he is doing with BOINC ;-)
(Rom is the release manager and a major developer of the platform)
So he'll know what works and does not, maybe fpops doesn't work well in rosetta (or maybe fpops is the internal benchamrk ;-))

SIDE/
As for the optimised app thing,
Well that's when you boinc credit system breaks down, since it's fine for 'in the same project credit' but bad for 'cross project credit'
Take Einstein for instance, maybe everyone not using the optimised apps should actually be getting less and have all there credit adjusted so and thoose using it should get the standard boinc credit ;) Why should other project suffer from a project (any) poorley coded application and when someone improves it now recons they should get more than a project who put the time in to write efficient code.
/SIDE

---

Team mauisun.org

Why should other project suffer from a project (any) poorley coded application and when someone improves it now recons they should get more than a project who put the time in to write efficient code.

I agree with FluffyChicken, let me just add that the issue isn't only of a "poorly written app", but sometimes the nature of the app itself. Sometimes you can get the main code to fit into L2 cache which speeds things up immensely.

Other times, a science app needs to constantly access memory, in which case FSB speed is the most important (in which case a 2GHz and 3GHz P4 might actually get about the same amount of work done per unit of time).

So, a lot depends on the nature of the science app (it also depends on developer skills ofcourse, as akosf has shown in the case of Einstein).

---

Best UFO Resources
Wikipedia R@h
How-To: Join Distributed Computing projects that benefit humanity