Hello,

My name is Rocco Moretti, and I'm a postdoc in the Baker Lab.

Chances are you might see something a little different when you look at your screensaver in the near future. The jobs I'm running on Rosetta@home aren't structure prediction, or protein-protein interaction design, but protein-small molecule interaction design. (These work units are tagged with "LigDes", standing for "ligand binding protein design") So in addition to the normal protein chain in the viewer, you'll also see a blobby representation of the small molecule.

With the current set of work units, we're looking at redesigning a genetic regulator to recognize different small molecules. All organisms must control their gene expression in response to different molecules in their environment - to turn on genes to take advantage of a new food source, for example. There's a number of different ways of doing this, but one of them is for proteins to recognize the small molecules, bind to them, and then have that bound protein go on to regulate genes. For example, the commonly used LacI repressor binds to DNA in its un-liganded state, turning off the genes it's bound to. When it recognizes and binds its small molecule, this changes the protein enough so that it no longer binds to DNA, allowing the gene it was bound to to be expressed. The LacI protein functions as a switch, effectively turning on genes only in the presence of small molecules.

What we want to do is take such a naturally occurring protein regulator and change it so that it binds not to its native small molecule, but to another, different small molecule. The hope is that this way we could create a set of different gene switches which are responsive to different small molecules. This should be of benefit to the field of synthetic biology, allowing for the control of multiple genes with multiple different small molecule regulators.

We're actually doing this research in collaboration with researchers in George Church's lab at Harvard, who have a way of "easily" turning large numbers of protein sequences in the computer to actual proteins in the test tube. As these proteins will directly control gene expression, it's very simple to test a large number of different variants very rapidly. That's where you come in - the number of variants we're looking to test is much larger than we typically produce locally, but should be a cakewalk for Rosetta@home. The current plan is that almost all of the designs that we get back from Rosetta@home will be directly tested in the laboratory.

Hello,

My name is Rocco Moretti, and I'm a postdoc in the Baker Lab.

Chances are you might see something a little different when you look at your screensaver in the near future. The jobs I'm running on Rosetta@home aren't structure prediction, or protein-protein interaction design, but protein-small molecule interaction design. (These work units are tagged with "LigDes", standing for "ligand binding protein design") So in addition to the normal protein chain in the viewer, you'll also see a blobby representation of the small molecule.

With the current set of work units, we're looking at redesigning a genetic regulator to recognize different small molecules. All organisms must control their gene expression in response to different molecules in their environment - to turn on genes to take advantage of a new food source, for example. There's a number of different ways of doing this, but one of them is for proteins to recognize the small molecules, bind to them, and then have that bound protein go on to regulate genes. For example, the commonly used LacI repressor binds to DNA in its un-liganded state, turning off the genes it's bound to. When it recognizes and binds its small molecule, this changes the protein enough so that it no longer binds to DNA, allowing the gene it was bound to to be expressed. The LacI protein functions as a switch, effectively turning on genes only in the presence of small molecules.

What we want to do is take such a naturally occurring protein regulator and change it so that it binds not to its native small molecule, but to another, different small molecule. The hope is that this way we could create a set of different gene switches which are responsive to different small molecules. This should be of benefit to the field of synthetic biology, allowing for the control of multiple genes with multiple different small molecule regulators.

We're actually doing this research in collaboration with researchers in George Church's lab at Harvard, who have a way of "easily" turning large numbers of protein sequences in the computer to actual proteins in the test tube. As these proteins will directly control gene expression, it's very simple to test a large number of different variants very rapidly. That's where you come in - the number of variants we're looking to test is much larger than we typically produce locally, but should be a cakewalk for Rosetta@home. The current plan is that almost all of the designs that we get back from Rosetta@home will be directly tested in the laboratory.

Brilliant! R@H is the most cutting-edge and least effort I've ever had to put into lab work! Will be great to hear how well R@H calculated bindings match the real-life versions. This work sounds similar in principle to the the Find-a-drug project quite a few of us used to crunch for six to eight years ago, although I imagine there have been huge strides on all fronts since then ;)

---

Hello there,

thank you for your detailed description. You may not see all the feedback that you should get but many people are pretty involved in the scientific side of BOINC and distributed computing in general, especially including projects like Rosetta@home where the science is a key appeal. However, there is some stir, not only on this site.

Several months ago I have put a brief summary of your post in Polish for my teammates in BOINC@Poland (link to the B@P wiki). I think you have already guessed that 1. not all the volunteers use the fora of the projects (but some of them heavily use the teams' communication channels) 2. not everyone's command in English is good enough to easily comprehend even popular-scientific texts in biology. This is why sometimes the volunteers act as a kind of "broadcasters" and "preachers" of the projects. :)

Now, as roughly 6 months passed, I would like to kindly ask you for some summary / follow-up post about your successes, failures and general work so far. I am quite sure that many people are interested in your project(s) and achievement(s) and more information would attract more volunteers.
Moreover, I don't need to mention that I would be pretty happy to provide my teammates a yet another reference point and a summary. :) I think it is very encouraging for many people to participate in distributed computing efforts.

So please, write us some updates if you find some time. :)

Best Wishes and happy crunching for everyone,
greetings from Warsaw,
a.m.
BOINC@Poland

Now, as roughly 6 months passed, I would like to kindly ask you for some summary / follow-up post about your successes, failures and general work so far.

Unfortunately not much "headline" level news to report. Such is science at times.

The collaborators who are doing the experimental testing of the designs have been working on improving the assay, so that we can get a clean "good binder"/"bad binder" readout once we start testing the synthesized proteins.

Another issue has been that the high-throughput gene synthesis technique we anticipated using isn't up to the task of synthesizing the full binding protein. (Specifically, the region over which there's variation is too large to synthesize in a single piece.) We've come up with a scheme where we'll break the binding protein gene into several regions, select the most common sequences for each piece, synthesize each of those shorter regions separately, and then randomly combine (in a test tube) those sections. Hopefully this will overcome the synthesis length limitation.

Thanks for your continued interest in the project! I'll continue to update you whenever there's more information to report.

Hi,

thank you for your fast response. I understand that the things are not always as easy as we thought they would be.

If I understand you correctly, everything what was reasonable to crunch for your project on this stage has been crunched and now we are waiting for the validation in an actual wet lab. Your collaborators from George Church's lab need to iron out two things: 1. a complete procedure of validating the computed results (as you wrote "good binder vs. bad binder") and 2. a proper method to actually synthetize the proteins to test them.

In such a case I should wish you good luck with physical proteins and many exciting computational projects. :)

Best regards,
a.m.

Finally, several years later, I can finally give an update on this project.

We were able to take the massive numbers of designs which you were able to generate (and truly, there were a lot - at one point we even had disk space issues with storing all of them), and examine all the mutations which were suggested. The initial plan was to make and test the designs produced by Rosetta@home as-is, but due to limitations in the gene synthesis platform in the Church lab, we had to break up designs into parts (based on region of the binding site) and make consensus designs based on those parts.

But with the processed designs in hand, our collaborators in the Church lab were able to express and test them for effectiveness in small-molecule-responsive gene regulation. From these computational designs, we were able to get workable variants for three of the four targets run on Rosetta@home, including sucralose. (The gentiobiose designs mentioned in the paper were not run on Rosetta@home.) Using standard (non-computational) molecular biology techniques, we were able to further optimize a protein to specifically bind sucralose - which also shows a response towards sucrose which is comparable to how the wild type protein responds towards the normal IPTG ligand.

The scientific paper describing the results is in this month's issue of Nature Methods: Taylor, et al. (2016) "Engineering an allosteric transcription factor to respond to new ligands", Nat. Meth. 13, 177–183. (pdf) And here's a less technical write-up on phys.org.

Thanks again to everyone who donated their computer time for Rosetta@Home, and who made this work possible!