Design of protein-protein interfaces

Message boards : Rosetta@home Science : Design of protein-protein interfaces

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next

AuthorMessage
Aegis Maelstrom

Send message
Joined: 29 Oct 08
Posts: 61
Credit: 2,137,555
RAC: 0
Message 64846 - Posted: 7 Jan 2010, 11:39:19 UTC

Great News and please, keep us updating! :) It should help to encourage us, the readers, and further our teammates, friends etc. to bigger involvement in R@H.
ID: 64846 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 65045 - Posted: 19 Jan 2010, 19:37:42 UTC

Hello,

I wanted to give you a brief update that I've received many excellent models of hemagglutinin binders from your computer! From the very large collection of designs that your computers produced I've selected 15 that we will start testing over the next few weeks. 15 is a very large number, but we want many many more so expect more such simulations in the near future.

Many thanks! Sarel.
ID: 65045 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
taw

Send message
Joined: 27 Jan 10
Posts: 1
Credit: 6,277
RAC: 0
Message 65125 - Posted: 27 Jan 2010, 0:57:20 UTC

Hi all,

my name is Tim Whitehead and I've been a postdoc in the Baker lab for about a year. My projects deal with the prediction and design of protein-protein interfaces. In this thread, Sarel has explained the design problem for protein-protein interfaces. Progress in this area can drive the next generation of treatments against pathogens like Mycobacterium tuberculosis or Influenza.

I have been working with Sarel on the design and experimental characterization of designs that target disparate proteins on both pathogens. We have a couple of successes thus far, which we are extremely excited about, but need to have better and better designs to show the validity of our design approach.

In this thread, Sarel has explained why and how we are targeting Influenza by designing a protein inhibitor of a protein that is displayed on the surface of the virus.

For Tuberculosis, our task is slightly more difficult. TB is caused by a bacterium called Mycobacterium tuberculosis. The bacteria is encased in a waxy coating of mycolic acid, which hurts our body's ability to expell the bacterium (for more information, see http://en.wikipedia.org/wiki/Mycobacterium_tuberculosis).
Instead of designing an enzyme that could degrade the coating, we are targeting a protein that is involved in the `synthesis` of mycolic acid. You may have already seen BOINC descriptions that say "docking of designs against 2CGQ" - 2CGQ is the protein that we are trying to inhibit. I am hoping that your contribution from ROSETTA@HOME will give us the increased quality of our designs that we need. Thanks for the help!
ID: 65125 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
strauch

Send message
Joined: 27 Jan 10
Posts: 1
Credit: 115
RAC: 0
Message 65126 - Posted: 27 Jan 2010, 6:34:38 UTC

My name is Eva-Maria Strauch and I am one of the post-docs in the Baker lab working on the design of protein-protein interactions. I am very excited about having the possibility to run on R@H, thanks to your generous contributions! Protein design requires a lot of sampling since there are so many many different possibilities to be considered carefully. This will work out now faster with your help!

I am currently designing proteins to bind to a model system which we use to calibrate our computational protocols. This particular target is the back end of a human immune-system antibody (Fc fragment, which is what you will see under the work unit descriptions http://en.wikipedia.org/wiki/Fc_region). The cool thing about Fc is that nature provides us with several proteins that bind to it, hence it taught us very important lessons on how to design possible binding proteins to it. We now want to see whether we've figured this one out and can mimic nature's ability to come up with binders towards this target. Success will further contribute to the optimization of our protocols to generate several protein-based inhibitors to attack cancer and infectious diseases. Questions we want to answere with this design project are: How well do we perform in comparison with nature? and if we understood all the information we got out of those native interactions, are we able to produce several de novo binding proteins that mimic the atomic interactions seen with native interaction Fc has with different proteins? This project will contribute to our understanding of how proteins recognize one another. Promising designs will be experimentally verified and also further characterized to see how well we performed, and what was possibly missing that we should include into our computational design protocol.
Thanks again for your contribution!
ID: 65126 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Michael G.R.

Send message
Joined: 11 Nov 05
Posts: 264
Credit: 11,247,510
RAC: 0
Message 65129 - Posted: 27 Jan 2010, 16:11:29 UTC

Thanks for the update, very interesting stuff. Keep up the good work!
ID: 65129 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jcorn

Send message
Joined: 27 Jan 06
Posts: 6
Credit: 198,437
RAC: 0
Message 65206 - Posted: 4 Feb 2010, 19:05:22 UTC

Hi boincers,
Since I last wrote, Tim Whitehead (username taw) and I have spent a lot of time testing the anti-IL23 designs that you made. So far we have a promising hit, but in science as in life, the results are somewhat ambiguous. I'll keep everyone updated when things are more concrete, but there's hope that your number-crunching has made a bona-fide binder for IL23!

I've also started designing against a new target, called Insulin-like Growth Factor 1 (IGF1). IGF is a central regulator of the speed of cell division: many tumors have been linked to having too much IGF around. Interestingly, the when an organism's cells are globally told to divide more slowly, by inhibiting IGF, the animal actually lives longer! As they say, the candle that burns twice as bright, burns half as long. In fact, since IGF is closely related to insulin and energy metabolism, some people have theorized that the increases in life span from caloric restriction are actually a side-effect of crosstalk with IGF signaling. So anti-IGF therapies hold promise as both anti-cancer agents and as anti-aging therapies.

I'm very excited to see what designs you can come up with, and everything is in place to test them in the lab. Look for workunits that start with "igf", eg - "igfhum" or "igffn3". Thanks for your help!
ID: 65206 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jcorn

Send message
Joined: 27 Jan 06
Posts: 6
Credit: 198,437
RAC: 0
Message 65207 - Posted: 4 Feb 2010, 19:06:49 UTC

For those who'd like more information on IGF-1, check out the wikipedia page

http://en.wikipedia.org/wiki/Insulin-like_growth_factor_1
ID: 65207 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Michael G.R.

Send message
Joined: 11 Nov 05
Posts: 264
Credit: 11,247,510
RAC: 0
Message 65216 - Posted: 5 Feb 2010, 20:07:42 UTC

Very exciting news, Jcorn. Thanks for the update. Am particularly excited about IGF research because I've read a bit about it lately. Seems very important in a lot of things that go right and wrong with metabolism.
ID: 65216 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
rickard

Send message
Joined: 12 Feb 10
Posts: 1
Credit: 145,915
RAC: 0
Message 65330 - Posted: 15 Feb 2010, 18:15:03 UTC
Last modified: 15 Feb 2010, 18:20:22 UTC

Hello,

my name is Rickard and I'm a visiting biotech student from Uppsala University (Sweden). I'm involved in a project aiming at creating proteins that bind to lysozyme. I'm being supervised mainly by Sarel, but I'm also getting help from the other postdocs in the lab. This work is the final part of my education, so it is great to be working in such an amazing group of people on this extremely interesting project.

While most of the others in the group are trying to design binders to therapeutically relevant targets, I am not. The protein I am working on, lysozyme, has for a long time been a favorite model protein among scientists. The reason is that it is very easy to work with, so there is loads of experimental data available. Of particular interest to us is the relatively large number of solved structures of lysozyme complexed with other proteins. This is extremely useful since we can learn a lot from looking at them. Nature has solved the binding problem in many different ways. It will be interesting to see if we can do something similar. By choosing an easy-to-work-with protein we make the biochemistry part of the project a little bit easier, and we are able to focus on testing and improving our design methodology. I would like to say that lysozyme, despite not being a pharmaceutical target, is an interesting protein to make a binder for. One reason is that lysozyme, just like many therapeutically relevant proteins, is an enzyme so if we can get a binder that inhibits its activity then it is likely that we can design inhibitors for other enzymes as well. At this point in time it is also the only enzyme that we are targeting in the lab.

Designing lysozyme binders with the humongous computational power that you are providing is very exciting! The results that Sarel, Jacob and the others have got from you have been very promising so I'm very much looking forward trying it out.

Thank you for helping out!
ID: 65330 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mad_Max

Send message
Joined: 31 Dec 09
Posts: 208
Credit: 24,633,259
RAC: 16,411
Message 65551 - Posted: 14 Mar 2010, 14:54:47 UTC
Last modified: 14 Mar 2010, 14:56:07 UTC

I got a strange WU which generates 10000 decoys (yes, ten thousand).
And does not show any graphics, except the 1st original(starting) model.
Link: fcDE-W3b_1dAl_1zzo_ProteinInterfaceDesign_11Mar2010_18648_49_0

This is a new type of protein-protein interfaces algorithm, which allows to discard "unpromising" model even more early than it was before?
Or is it a just fault in the program?
ID: 65551 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
strauch
Volunteer moderator
Project developer
Project scientist

Send message
Joined: 15 Mar 10
Posts: 7
Credit: 40,011
RAC: 0
Message 65556 - Posted: 15 Mar 2010, 5:56:27 UTC - in response to Message 65551.  

I got a strange WU which generates 10000 decoys (yes, ten thousand).
And does not show any graphics, except the 1st original(starting) model.
Link: fcDE-W3b_1dAl_1zzo_ProteinInterfaceDesign_11Mar2010_18648_49_0

This is a new type of protein-protein interfaces algorithm, which allows to discard "unpromising" model even more early than it was before?
Or is it a just fault in the program?


thanks for pointing this out! That does sound weird. We are looking into what is going on. The WU did produce some valuable decoys though! but we are taking those WU off till we figure out what could be the cause of this.
ID: 65556 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Tulip

Send message
Joined: 21 Jul 08
Posts: 1
Credit: 2,363,164
RAC: 0
Message 65569 - Posted: 17 Mar 2010, 2:12:48 UTC

Hello.

A Question for Administrator`s /scientist`s:

Sorry but i am realy not sure how to ask :p Anyway...

What is the chanse that rosetta@home can investigate/"do a research" on
SMN1 and SMN2 protein?

http://en.wikipedia.org/wiki/SMN1 and http://en.wikipedia.org/wiki/SMN2

If not.. Who do you recommend i can contact in distributed computing "family" to get some numbers crunched on SMN1 and SMN2.

Sorry my realy bad English.
ID: 65569 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 66178 - Posted: 17 May 2010, 19:22:09 UTC
Last modified: 17 May 2010, 19:25:37 UTC

Hello,

As you've seen on David's thread a design from your Rosetta @ Home contributions is a tight binder of influenza's hemagglutinin protein making it a candidate to serve as an inhibitor and hopefully a therapeutic. This is exceptionally good news. Many thanks to all of the participants!

This news is making us all very excited of course and we are starting to think about the next steps. One promising avenue for future research is to use this computational technology to design specific inhibitors of proteins that are important regulators of cellular activity but where experimental research has been stymied by the lack of specific, non-toxic effectors. Let's take a few steps back and try to see the broader picture. Our understanding of molecular biology progresses through the identification of key players and systems in the cell and then by modulating these effectors and observing the results on cell viability and function. This process parallels the concept of reverse engineering of mechanical and electronic systems.

A major tool in this process involves the use of genetic knockouts. These are mutations to genes that disable a protein's function. Using this technique myriad gene functions have been described, including genes that are important tumor-suppressors or promoters, metabolite importers and regulators of cell size, form, and fate. But, gene knockouts are not very subtle tools and they often have far-reaching consequences on cell viability, making it difficult to interpret the results of the knockout experiment. For a beautiful and accessible description of the parallels between reverse engineering and molecular biology and where the difficulties lie, see "Can a Biologist Fix a Radio?" (Biochemistry (Moscow), Vol. 69, No. 12, 2004, pp. 1403-1406; protein.bio.msu.su/biokhimiya/contents/v69/pdf/bcm_1403.pdf).

One way to deal with these difficulties is by using specific inhibitors that are invoked at specific times to block a certain protein function rather than pulling the brakes on a protein from the cell's inception. The ideal inhibitor would selectively modulate one interaction without affecting other interactions, thus allowing a subtle study of the significance of interactions in the cell. Thus specific inhibitors, where they have been discovered and deployed, have yielded very important lessons on the function of proteins that are crucial in many cellular signaling pathways, to give but one example. But, these inhibitors are rare and very difficult to identify. We think that similar to the influenza binder, we should be able to use computational protein design to generate novel and highly specific protein inhibitors of important cellular players and help in elucidating their functions.

Over the next few weeks we will launch new jobs on Rosetta @ Home that target proteins where an inhibitor could help make progress in our understanding of cell biology. As we prepare each of these targets for Rosetta @ Home we will describe the rationale for pursuing them in this thread.

One of these targets is RhoA. This protein belongs to the family of oncogenic Ras proteins which are involved in cell division and regulation. RhoA is known to be a major player in cancer. As mentioned above, knocking out RhoA causes many different problems for cell viability. It is easy to understand why if you merely look at the number and variety of its interaction partners (see, http://en.wikipedia.org/wiki/RHOA)! By generating a specific inhibitor of RhoA we hope to enable probing the effects of its modulation at specific points in the living cycle of the cell and under different conditions -- studies that are nearly impossible to perform today.

Finally, I'd like to thank all of the participants in Rosetta @ Home for the immense boost in computational throughput over the past few weeks! Without the 40% increase in computational power it would have been impossible for us to work on these targets in parallel to the needs of the group who are working on CASP! Please keep these contribution levels!
ID: 66178 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
strauch
Volunteer moderator
Project developer
Project scientist

Send message
Joined: 15 Mar 10
Posts: 7
Credit: 40,011
RAC: 0
Message 66323 - Posted: 25 May 2010, 18:14:16 UTC

I just wanted to update you on another project I have been working on and I am running on rosetta@home.

Enterohaemorrhagic and enteropathogenic E. coli strains (like O157:H7, http://en.wikipedia.org/wiki/E._coli_O157) are some of the most important food-borne pathogens in North America, Europe and Japan, and contribute worldwide to pediatric morbidity and mortality. You might have heart about the outbreaks in the States that involved green spinach, iceberg lettuce or ground beef.

In order to cause diarrhea, these bacteria attach to the intestines before they get into your body. For that they have developed a special anchor-hook system to stick to the cells of our guts just like Velcro. I am targeting this molecular mechanism to produce protein-based inhibitors that prevent bacterial attachment and thereby reduce or prevent their colonization of our guts.

Thus far, traditional approaches to neutralize this particular interaction have not been very successful. Neither antibodies nor small peptides have been efficiently produced that target the interface of this molecular
machinery. We think that our new methodology could facilitate the development of a new therapeutic against these bacteria. You will see several jobs that are called "design against intimin" or such.

Thanks for supporting us with your computer time! It is a great help!

Eva

ID: 66323 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 66469 - Posted: 5 Jun 2010, 4:28:43 UTC

We've started to design proteins against another exciting target called MDMX. MDMX is known to interact with one of the key tumor-suppressor genes, p53. p53 has such a central role in physiology that it has been called "guardian of the genome". MDMX inhibits p53 and causes its degradation through a complex mechanism. But, the specific role of MDMX in p53 inhibition is something of a mystery because a close homolog of MDMX called MDM2 is involved in very similar interactions with p53. The difficulty in deciphering the specific roles of MDMX lie in the fact that to date a specific inhibitor of MDMX has not been identified. That is, inhibitors of MDMX inhibit MDM2 to the same extent so that perturbing MDMX alone (and vice versa) is very difficult.

Our goal in this project is to design an inhibitor of MDMX that would not interact with MDM2. We plan to take advantage of subtle differences between the two proteins in order to design a protein that will be highly specific to MDMX. Hopefully with this protein in hand, research into the role of MDMX will gain a new and important tool.

Over the next few weeks I will submit jobs with the word MDMX in their title to Rosetta @ Home. I'm very excited to see what comes out!

In the meantime, here's some further reading for those who are interested in the intricacies of this important system:
http://en.wikipedia.org/wiki/Mdm2
http://en.wikipedia.org/wiki/P53
http://en.wikipedia.org/wiki/MDM4 (MDMX is also known as MDM4)
ID: 66469 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
P . P . L .

Send message
Joined: 20 Aug 06
Posts: 581
Credit: 4,865,274
RAC: 0
Message 66481 - Posted: 6 Jun 2010, 0:34:55 UTC

Hi Sarel.

That didn't take long, i've gotten one of your new tasks this morning.

MDMX_3jzp_1o6d_ProteinInterfaceDesign_4Jun2010_21359_52

ID: 66481 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mad_Max

Send message
Joined: 31 Dec 09
Posts: 208
Credit: 24,633,259
RAC: 16,411
Message 66514 - Posted: 8 Jun 2010, 1:44:14 UTC

Only I went to ask about a new type of tasks, which drew my attention... And here is a ready description posted already. Excellent speed.
But the question still is: it is simple (short) protein? (But in wiki i found info about 490-amino acids, it is not short i think) Or improve of algorithm? This type of job (MDMX_ *) generates much more decoys for the same time, compared with other protein-protein tasks i saw before. From a few hundreds to 1200 decoys in only 2 hours (7200 sec) tasks at mid-level processor. I already had a lot of these tasks, that's only 3 for example:

https://boinc.bakerlab.org/rosetta/result.php?resultid=343957927
https://boinc.bakerlab.org/rosetta/result.php?resultid=344057490
https://boinc.bakerlab.org/rosetta/result.php?resultid=343943380

And judging by the normal ratio between Claimed and Granted credit is not a feature of my computer, but it happens at all.
ID: 66514 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 66515 - Posted: 8 Jun 2010, 3:31:10 UTC
Last modified: 8 Jun 2010, 3:31:57 UTC

Some of the newly developed protocols have phases which produce models rapidly, even for larger proteins. So entire lines of tasks are created which can scour the landscape and locate the major features of the terrain. Then this information can be used to create additional lines of tasks which focus on the points of highest interest.

You are correct, these tasks run "fast" on everyone's machine, as compared to other typical tasks anyway. So the credit per hour of CPU time should be in line with other tasks. Since the models are relatively easier (then models for other tasks) for a machine to produce, each is granted less credit.

I believe those were the main questions you had. If I missed it, let me know.
Rosetta Moderator: Mod.Sense
ID: 66515 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 66522 - Posted: 8 Jun 2010, 17:09:33 UTC - in response to Message 66514.  

Thanks for your feedback! Mod.Sense's replies on the conformational sampling is very accurate. With regard to the size of the protein, you are right that MDMX is a much larger protein than what we are modeling in these design trajectories. What we are interested in is the interaction surface of MDMX with p53. That is (thankfully) a very small domain allowing us to do things more quickly and efficiently.

Only I went to ask about a new type of tasks, which drew my attention... And here is a ready description posted already. Excellent speed.
But the question still is: it is simple (short) protein? (But in wiki i found info about 490-amino acids, it is not short i think) Or improve of algorithm? This type of job (MDMX_ *) generates much more decoys for the same time, compared with other protein-protein tasks i saw before. From a few hundreds to 1200 decoys in only 2 hours (7200 sec) tasks at mid-level processor. I already had a lot of these tasks, that's only 3 for example:

https://boinc.bakerlab.org/rosetta/result.php?resultid=343957927
https://boinc.bakerlab.org/rosetta/result.php?resultid=344057490
https://boinc.bakerlab.org/rosetta/result.php?resultid=343943380

And judging by the normal ratio between Claimed and Granted credit is not a feature of my computer, but it happens at all.


ID: 66522 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mad_Max

Send message
Joined: 31 Dec 09
Posts: 208
Credit: 24,633,259
RAC: 16,411
Message 66528 - Posted: 9 Jun 2010, 12:50:37 UTC

Thank you. Now it is clear from this acceleration comes. Due to the possibility of modeling is not all the protein completely, but only the most important (for the current study), part of it.

2 Sarel
And yet another question. I am not a scientist, so this may seem silly, but as a result of observing the process of calculations I had the idea.
It concerns the limit of 500 steps in modeling protein-protein including the last MDMX. It is clear that the shorter limit is introduced to speed up processing. And in most cases it seems to be enough - a graph of energy usually manages to "find minimum" for these 500 steps and more just jumping around him.

But periodically I see a model in which the energy almost continuously go down, but the simulation of the same breaks at the limit of 500 steps. (Obviously not a straight line, but with variations here and there, but the overall trend is absolutely clear - down) Though if given the additional steps would be found a configuration with a much lower energy. And since this model does not differ from the total mass, because its calculation was stopped before she could reach its minimum. Simply increasing the number of steps are not very effective way, because it increase the use of computing resources in times (or reduce the number of models with fixed resources), for 5% of models (about as much on my observations do not manage to reach the "saturation") is not effective. (Though perhaps this is the most valuable model, so that theoretically could be more valuable than the other 95%)

But what if instead of a fixed hard limit on the number of steps to use a dynamic limit is based on "delta energy" criteria? Ie compare the minimum energy found suppose the last 100 steps (for example, concrete value must be chosen experimentally), with previous values. If there is some improvement, then give the model additional steps, and so until the reduction of energy does not stop. The fixed limits (in particular number of steps) are also needed, but only as a acceptable framework - minimum (say cover those 500) and maximum (it should be big enough, but within reasonable limits. To finish on time "bad model", without waiting for the activation of watchdog).

Perhaps this idea has already been tried before? Then it would be interesting to know the results and why abandoned.
ID: 66528 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 . . . 6 · Next

Message boards : Rosetta@home Science : Design of protein-protein interfaces



©2024 University of Washington
https://www.bakerlab.org