Our Gates foundation grant to develop HIV vaccines started earlier this week (Aug 1), and it has been an exciting time as we are finally able to pursue this goal actively rather than just thinking about it! We have already ordered reagents to put together the first five designs, and there are many more in the pipeline. After this first round of designs is complete, and the designs are close to being sent to our collaborators to be tested as possible vaccines, a high priority will be to extend rosetta@home to design calculations (which shouldn't be difficult, as the same underlying rosetta source code is used) so that all of you can contribute to the second round of designs.

This morning I had to wake up early to go to a BBC radio interview which I think several of you participated in as well. The more media attention, the better for the project, but I'm really not very good at this kind of thing...

Meanwhile, CASP7 has only a few days more to go and we are waiting for the solved structures to be released so we can see how well we did. so far only a few novel structures have been released, our predictions for these are good, but not at atomic resolution, probably because the proteins are longer than those in our pre CASP tests and the search problem is correspondingly more difficult.

---

You are currently carrying out calculations for the CASPR structure refinement challenge. In this case, rather than having to predict the structure of a protein from the sequence information alone, we are given both the sequence of the protein and a starting model which is not too far away. The challenge is to refine the structure to closer to the actual true structure. In the landscape searching analogy, this corresponds to being told the valley the lowest elevataion point lies in, but not the exact location of thie point.

The refinement problem is similar in concept to the second "high resolution" stage of our standard prediction protocol. Bin uses essentially the same code to carry out these refinement calculations, except that some portions of the starting model are randomly rebuilt at a low frequency to allow a broader search arround the starting model. As you have undoubtably noticed, there is relatively little movement during the high resolution refinement protocol, and this random rebuilding broadens the search considerably.

---

It is an exciting time in the lab now as we are recovering from the craziness of CASP. While Bin and Rhiju are taking an incredibly well deserved vacation, the new HIV vaccine design project is starting to come into full swing. We now have computationally designed amino acid sequences for 15 potential vaccine candidates, and we will start the process of making them next tuesday; the first step is to synthesize genes which encode these proteins. We have also designed a whole series of novel enzymes which catalyze a wide variety of reactions, and are starting the gene synthesis process for these as well.
I'm particularly interested now in designing enzymes which destroy organophosphate compounds which are the key ingredients in many pesticides and nerve agents. On rosetta@home, we are carrying out calculations in which we are resampling regions of the landscape found to be low energy in initial sets of runs and we hope these will lead to significnat improvements in our abilities to find global minima.

I'm very sorry about some of the not nice things being passed about on the message boards, and I'm also sorry that my efforts to calm things down haven't helped, so I will be doing my communicating with the project solely through this thread for the next week. You are all making great contributions, and I ask people who have been annoyed by what has been said on one side or the other to try to think about the big picture and what we are all trying to accomplish together.

---

A couple more things:

First, I will be describing all of your efforts at a conference on Monday on grid computing:
http://www.opensciencegrid.org/events/meetings/consmeeting0806/agenda.html. (should I
show some excerpts from the message boards?). I will of course describe the great work you have all done together.

Second, To answer a question which came up on the boards--we will NOT be backdating credit totals. The new system will go into place early next week, adding on to the current totals.

Again, I thank you all for your efforts and the many people who have volunteered to help recently!

---

A BBC report on distributed computing is at http://www.bbc.co.uk/radio4/science/citizenscience.shtml
(I was interviewed, but haven't listened to it).

Longer segments on Rosetta@home were made recently by media groups in the US; I'll pass along the links as I get them.

I'm now preparing my talk for the grid computing meeting on rosetta@home; after I discuss the science and the results, I'm going to pick a few message boards threads tomorrow morning to illustrate the issues that come up--any suggestions on which ones to show? (this could be a good time to clean up recent posts)

---

I showed a bit of the rosetta@home web site, including the message boards, at the grid computing meeting this morning--I think this is the part of my talk that people understood the best. If people are interested all the talks from the meeting will eventually be posted on the web.

I'm working tonight on a manuscript with my former graduate student Rich Bonneau on some of the results from HPF1 done on the world community grid. We predicted structures for all the proteins in one of the best studied eukaryotic organisms--the yeast used to make bread and beer, and then integrated these predictions with other experimental data to assign 500 proteins of previously unknown structure to protein structural families. After this is done, we will start working on the report on the structures of human proteins also done in HPF1. These efforts used the low resolution version of rosetta (which is all we had several years ago when the HPF project started); I am of course excited about HPF2 which is using the protocol we have been improving on rosetta@home (I sent Rich and the collaborators at IBM the code last March) and should produce much more accurate models.

On the credits front: we have decided to use the average amount of time for producing a structure over all rosetta@home runs for a particular work unit to determine the amount of credit to be awarded for each structure produced for that work unit. so, for example, let us suppose that all rosetta@home computers on average took 1 hour to make 1 structure for a given work unit, and that this corresponds on average to 10 credits using the standard boinc accounting scheme. Then each computer gets 10 credits for each structure returned--a fast computer might be able to do 3 structures in an hour, and get 30 credits per hour, wheras my old slow laptop may require 2 hours to make a single structure so I would only get 5 credits per hour. I think everybody will be happy with this approach in the end, even though nobody may be very happy with it initially (I must emphasize that, contrary to some statements on the boards, no individuals or groups had any more influence on our strategy than any other, so I hope this issue can be laid to rest). I am sorry that the switch to the new system has generated so much conflict, I really didn't anticipate this, and I'm sorry that my attempts to calm things down only made things worse (again, I will post from now on only in this thread and only in this forum). In any event, we are quite set on the new credits plan, which we think will be better for everybody, and please hold off on comments or suggestions for two weeks or so until we all have a clear picture of how things are working--we do not need or want new suggestions at this point. David Kim will be posting the definitive description of the new system tomorrow on the boards.

A positive outcome of the recent animated discussions is that we have recruited a number of new moderator volunteers who will be introduced soon and who will be ensuring that the boards stay friendly (please help the project out by avoiding posting things likely to offend other participants!). Also, as I'm pretty much tapped out on corresponding with participants by emai (which could easily turn into a full time occupation!), after tonight if you have issues please contact the moderators via the boards or email and they will pass unresolved issues on to me.

OK--thats enough of this topic, my posts will be back on the new science discoveries track starting tomorrow!

---

Today was an exciting day for the group! In our vaccine design and enzyme design calculations, the end result is an amino acid sequence for a protein predicted to be a good vaccine or catalyst of a chemical reaction. The next step is to make a gene--a piece of DNA--that codes for the amino acid sequence. Due to advances in technology, rather than having to laboriously synthesize each gene in the lab, we can buy genes for any amino acid sequence for not to much from DNA synthesis companies, and we are lucky to be collaborating with a startup company in Boston called Codon who can make them for us quite cheaply. Today we ordered genes for 16 potential HIV vaccines, 15 potential new enzymes, and 4 potential new protien-protein complexes. I say potential above because our design calculations are not perfect, and we won't really know if these proteins act as designed until after we get the genes back in a month or so. Then we take advantage of modern molecular biology techniques to put the genes into bacteria where they direct the cells to make large amounts of the designed proteins. We can then separate the designed proteins from the rest of the stuff in the bacteria using a special tag we include in each of them that provides a good handle. Once we have the purified designed proteins, we can see whether they bind the desired antibodies in the case of the vaccine designs or catalyze the desired reactions in the case of the enzymes. In this way, we will learn about both the strengths and weaknesses of the rosetta design methodology, and hopefully have crated proteins that can have a very positive effect on the world!

As Hugo pointed out, we have not quite gotten the design methodology to the point we can run it on rosetta@home, but this should be coming in not too long as several people in my group are now focusing on this. Before this, look for protein-protein docking calculations where we are trying to predict the structures of the complexes between proteins which mediate much of the basic processes important to life. Chu Wang, a graduate student in the group, is close to having his docking methodology compatible with distributed computing, and we anticipate breakthroughs in this importnat problem as it also seems largely limited by cpu power.

Currently running on rosetta@home are the last of the casp tests on protein structure refinement (see the casp7 website) and tests of a general approach for estimating how much compute power is necessary to find the lowest energy structure for a sequence. I will describe the basic idea behind this approach in one of my next posts.

---

Tonight I want to describe the approach we are taking, in collaboration with several other research groups, to trying to cure human diseases caused by mutations in critical genes in cell populations that are capable of self renewal. An example of such a disease is sever combined immunodeficiency, which is described at http://www.scid.net/. The idea is that if we could correct a crippling disease causing mutation in for example the blood cell population or the immune cell population in even a very small number of cells, then these now normal cells could divide and eventually repopulate the body with healthy normal blood or immune system cells.

To accomplish this, we would need to target specific DNA sequences around the site of the disease causing mutations in the critical genes. We are doing this using the computational design methodology we described in the Nature paper earlier this summer that I mentioned in an earlier post. We are now designing enzymes designed to cut precisely in the genes responsible for SCID and other diseases. When we have succeeded in creating enzymes that cut specifically within these genes, and not in other parts of the genome, our collaborators will introduce these designed enzymes into mutant cells with a copy of the normal gene that doesn't have the mutation. Cells repair breaks in their DNA by copying from identical or near identical sequences elsewhere in the genome, and it is likely that this introduced DNA would be used to repair the break, in which case the mutation would be corrected. Of course, it would still be a very long road before such an approach could be used clinically, but it is an exciting road to be getting started on!

I'll be out of internet access for the next 9 days, so keep up the great work! Just before I left today Ben a graduate student in the group showed me some very exciting preliminary results from the rosetta@home jobs you have been doing; he can find very low energy low rmsd structures much more efficiently using information gleaned from a first round of searching than with our standard random search protocol. This was for one test protein, by the time I return you will have completed calculations using his new approach for a number of test proteins and we are excited to see the results!

---

The solutions for many of the CASP7 prediction problems many of you worked on May-July have now been released and we are currently comparing them to the predictions we submitted earlier to the CASP organizers. Rhiju has just posted on the "top predictions" page a side by side comoparison of the predicted structure for one of these proteins, target 299, with the actual structure. As you will see, it is quite close despite the considerable complexity of the structure. This was a very interesting target because Rhiju employed the full spectrum of approaches we have been devloping en route to making, with all of your help, this quite excellent prediction.

---

I am delighted to thank the four new moderators who are taking care of the rosetta@home message boards--they are doing a great job! I asked them to write a few sentences about themselves, and I'm posting these below. Again, I deeply appreciate their efforts on behalf of the project!



Mod.DE
I participated in some DC projects a few years ago (i.e. RC-64) but stopped.
February this year I found BOINC and later Rosetta where I’m feeling quite
happy at the moment. I’m kind of a computer geek but refuse to work in that
area (in order not to become a total geek) and try to finish my diploma
thesis in economics instead. I have a faible for all kind of social
interaction over the internet currently called "Web 2.0" but on the same
time think I should spend less time before my computer.

Mod.Canada
Some of the things that I am into are hunting, fishing and camping. I enjoy the "outdoors thing". I also like to tinker with DIY Audio circuits and hifi sound, along with computer mods of all kinds such as case mods. I do lots of overclocking, and have a real new found love of Computer language studies mainly being C++, and a little Perl.

I have also been battling Crohn's Disease for 6 years and have had several operations along the way. I have also come down with another life long disease called Avascular Necrosis which was caused by the large amount of steroids used to treat the Crohns disease. It is mainly looked at as a loss of blood flow to the bones, which in turn causes them to die over time.

Mod.Sense
I'm an armchair scientist that came project hopping, and found the insight into science and the high level of transparency to the research team to be compelling reasons to stay and crunch Rosetta full-time. I've stopped looking for another interesting project, and started looking for ways to help the Rosetta team achieve its goal of developing the science that will reinvent medicine, as we know it.

Mod. Tymbrimi
I've spent most of my DC time these last 4 years on F@H, but have joined teammates on other projects like D2oL and FaD. After Dr. David Baker posted an invitation for FaD participants, many of the team moved here.

I'm a fan of another infamous David. Mr. Brin. Even if he does talk about brushing mule's teeth.

It's still amazing to have started with the original release of programs with a sew on patch stating, "Boldly going where Angels fear to tread." in the 1980s, to actually helping perform scientific research with our spare cpu cycles.

---

We are just starting to model the structure of the fibrils which accumulate in Alzheimer's disease. Phil Bradley here has developed really exciting new methods which allow him to model both the folding of the protein and the assembly into fibrils simultaneously. We are collaborating with people who are making experimental measurements on the fibers, which provide constraints which will be very valuable for evaluating our models. I will keep you posted on these calculations, and hopefully you will see them running on your screen savers in not too long. (In addition to its medical importance, the simultaneous folding and spiral fiber formation is really neat to watch!).

---

Our first vaccine design calculations have just been sent out on rosetta@home! We are taking a number of different approaches to vaccine design; in the calculations that are just being sent out we are aiming to stabilize a portion of the HIV surface protein, GP120, in its active conformation so it can be used as a vaccine. We are keeping the portion of the protein that interacts with a cell surface receptor called CD4 constant, but redesigning the remainder to increase its stability. The lowest energy designs that are returned from all of your runs will be analyzed here, and the most promising of these sent to our collaborators at the NIH for testing as possible vaccines.
This first batch of work units is relatively small--look for work units beginning "PSH"; there will be a much larger batch next week and we will give them more informative names so you can tell what you are looking at.

---

Bin has just posted on the "top predictions" page a quite stunning example of success with the "high resolution refinement" method we have been developing. For this CASP target, there was a structure of an evolutionarily related protein that was already known. In such cases, the structure prediction problem is called "comparative modeling" and the challenge is to start with the evolutionarily related structure and refine it towards the true structure (the basis of comparative modeling is the empirical observation that evolutionarily related proteins almost always have similar (but not identical) structures).

For this prediction, Bin started with the red model, and used the high resolution refinement protocol from the second step in the ab initio prediction protocol (the part which is less exciting to watch because most of the action is with the sidechains which we don't display on the screensaver). He used your computers to carry out very large numbers of independent refinement runs, and chose the lowest energy structures to submit to CASP. As you can see, his submission, the green model, is very much closer to the true structure (in blue) than the starting red model.

These changes may seem subtle, but from the point of view of designing drugs to interact specifically with a protein structure, and understanding precisely how a protein machine works, this level of accuracy really is critical. Needless to say, we would be delighted if we could consistently refine models to this level of accuracy!

---

Laura has just posted a new version of her rosetta@home video--see her post in the rosetta video thread below. If you have a chance, take a look and post in her thread any suggestions you have for improving it.

---

Graduate student Chu Wang has added sidechains to the screensaver, so you will soon be able to see them flickering around during the fullatom refinement stage of the calculations. Chu has also made his flexible backbone protein-protein docking approach compatible with BOINC, so you will soon be seeing pairs of proteins searching out the lowest energy docked conformation on your screen savers. This is a very important problem as much of biology depends on precise and specific interactions between pairs of proteins.

---

Rhiju has just posted another ab initio prediction made for CASP7 using all of your computers. This is an all beta sheet protein, which have traditionally been the most difficult to predict because the interactions involve residues separated by long distances along the chain. If you compare the predicted structure for this protein, CASP7 target t0316, to the recently released native structure, you will see they are very similar. The superposition with the sidechains of both the predicted and native structure show this particularly clearly. Thanks to bwpow and JVMerlino for finding these excellent predictions!

---

David Kim and Stuart Ozer from Microsoft Research have been working hard for the past two months to help give participants more feedback on their results. The results of their efforts thus far are now available--as David describes below, you can now see both the overall results on each work unit, and the contributions you individually, or your team has made. David has started a thread in the Science section which addresses questions which are coming up--we are delighted at the positive response so far!


from David's post:
You can now view energy vs rmsd plots for active work units. To view your results, click on the "Results" link under "Returning participants" on the home page. To view results from the top users, hosts, and teams, click on the "Rank" numbers on the respective leader lists. The data gets updated daily.

We'd like to thank Stuart Ozer from Microsoft Research for helping us develop this feature.

---

Lots of exciting stuff these days!

Rhiju just posted a spectacular CASP7 prediction--see the top predictions page. thanks to the contributors!

Oct 14, 2006 - CASP7 target T0283
sarha1 (Team Czech National Team)
XS_DDTUNG (Team XtremeSystems)
cupra (Team XtremeSystems)
rtroll
XS_vapb400 (Team XtremeSystems)

I hope you have all had a chance to look at the results your computer has been generating (accessible from thre "results" bar in the "returning participants" section of the page. we really like this reporting tool, and are starting to use it for many of our analyses.

You will have seen by now on your screensavers the addition of the protein sidechains during the "fullatom relax" stage of the simulations. This gives you a more complete picture of the "3 dimensional jigsaw" nature of the protein folding problem, where the challenge can be viewed as getting all the pieces of the puzzle to fit together perfectly with no holes.

Our dream now is to make rosetta@home interactive, so you can move the chain around if you see a possible way to solve the puzzle. we are talking with colleagues in the CS department here who are experts on video games about how to approach this. eventually you could imagine designing proteins to cure diseases for fun and relaxation--I think it is possible that it could be made as engaging as a standard computer game. (we haven't thought about the ramifications for credits, but if you can guide the simulation yourself, you should get a higher score for finding lower energy solutions ... . but we won't have to cross this bridge for quite a while!).

---

Welcome to the new participants--we are delighted at the increase in users over the past few days! please tell all your friends and relations!

I posted a brief explanation of the "top predictor" award we post every day on the home page in answer to a question on the number crunching boards; I'm copying the question and my response here for people who might not see it there:

question

"I was absolutely thrilled to see that my team has helped the Rosetta@home project enough to be chosen as the predictor of the day. I was hoping for this for a few months now as our output has gradually increased. This is very exciting!

I am just wondering.. how does the predictor of the day relate to the top predictions, as seen on https://boinc.bakerlab.org/rosetta/rah_top_predictions.php?

Also, where can I see the actual structure which has the top prediction? It is no longer listed on https://boinc.bakerlab.org/rah_results.php?TeamID=1291.

Thanks and best wishes to all! "

my response:

Thanks and congratulations!! As Feet1st explained, you found the lowest energy structure for the indicated CASP7 work unit for CASP target 354. During CASP we experimented with a number of different strategies for each target; the different work units for the same target protein represent the different strategies. The participants acknowledged in the "top predictions" section are the people who found the lowest energy structures over all the different strategies (I can't tell whether your model was one of these because we haven't put the data together yet).

Your result, like those of the other "top predictors" was important because it gave us important feedback on the strategy being tested. target 354 I think was one where the predictions are really good--your model may have been one of the top 5 models submitted!

to view the predictions that you have recently made, you can follow the directions on the home page. to view the distributions of energies and rmsds for your predictions and those of your team, follow the "view your results" link from the home page.

we don't currently have a mechanism for you to retrieve the structures for predictions you made in the past because of the large amount of data we would have to have accessible, but if there is interest we might be able to do something along these lines.

---

Bin has posted in the top predictions section the rosetta@home based blind prediction for the CASP7 target 330 structure. This is an impressive illustration of the all atom refinement procedure we have developed to improve low resolution models. In this case, starting from a rough model based on a protein with a related sequence and known structure, shown in red, Bin carried out large numbers of independent refinement runsl with all of your help. The lowest energy structure he found is shown in green and you can see it has moved very much closer to the true structure (blue) which of course we did not know when the predictions were made.

Thanks to the following users who contributed particularly low energy models for this target:
Marko (Team Serbia - The Wild Bunch)
WindForce (Team XtremeSystems)
Ian_D
pxee (Team Poland Null-Zero Team)
csbyrosetta (Team SETI.Germany)

---