Hi Rosetta researchers,

I was curious about what the DRH_Curve jobs are modeling? There seems to be an endless stream of them for the past month or two.

Do these jobs all belong to one person? Are they all looking at a particular protein / family of proteins, or a particular type of situation or..? Just curious.

---

**38 cores crunching for R@H on behalf of cancercomputer.org - a non-profit supporting High Performance Computing in Cancer Research

Do these jobs all belong to one person? Are they all looking at a particular protein / family of proteins, or a particular type of situation or..? Just curious.

We seem to get very little information about what we are doing. Folding, for example, gives a description of each type of work unit. I would hope that they answer your question.

---

I would hope that they answer your question.

+1

---

I was curious about what the DRH_Curve jobs are modeling? There seems to be an endless stream of them for the past month or two.

Are months that, on android devices, we are crunching "cispro_backbone_".
What is it?

---

And, as usual, no answer....

---

Hello everyone,

My name is Jacob O'Connor and I'm one of the graduate students in the Baker lab. I don't know about the DRH_Curve but all of the cispro labeled job runs are likely mine. I'm not sure how much everyone here knows about the details of proteins so please forgive me if I re-explain something you already know.

Every protein is made up of amino acid building blocks linked into a chain. These building blocks consist of three major components. A 'N-terminus', the 'R-group' or side-chain, and a 'C-terminus'. To link these building blocks together a chemical bond is created between the N-terminus and the C-terminus of adjacent amino acids, replacing both with a peptide bond. This peptide bond has partial double bond character due to resonance stabilization, which effectively means it will always have an angle of near 180 or 0 degrees (lying flat). These two conformations are referred to as either 'cis' or 'trans' peptide bonds. In most cases the 'trans' conformation is much more favored because in the 'cis' conformation the side chains of adjacent amino acids bump into each other while in the 'trans' conformation they do not. That means when we model proteins we can effectively ignore cis peptide bonds and treat them all as trans, which is useful as it means we have one less degree of freedom we need to optimize when making a model.

However there is one notable case where this is not true. The amino acid proline has a ring structure that results in its side chain bumping into its neighbors regardless of if its in the 'cis' or 'trans' state. So while the trans state is still favored its much less so for proline than for any other amino acid. As such the majority of times when cis bonds due occur in natural proteins its at a proline position - hence 'cispro'

http://www.cryst.bbk.ac.uk/PPS95/course/3_geometry/peptide2.html

So why am I making so many jobs that have cis-pro? Three reasons. Firstly I work with a group of researchers in the lab who are really interested in studying very small proteins called peptide macrocycles. Because these peptides are so small (only 6-14 amino acids usually) and they are under chemical constraints the normal rules of protein structure don't apply quite the same way. We rely heavily on prolines to give our designs structure and so its important for us to model these prolines in both the cis and trans state to make sure the state we want is the lowest energy structure.

Even so, what I'm doing is a little bit different than what others working on peptide macrocycles are doing. Because I am making peptide macrocycles that are supposed to have cis peptide bonds in them. I am interested in that for two reasons. Firstly, our group is always interested in making new shapes with proteins and the shapes you can make with cis bonds are different than the ones you can make with trans bonds. And we can only ever really reliably make cis bonds in these peptide macrocycles, because of their unique chemical constraints and because they are small enough that adding the extra degree of freedom at the cis bond wont completely explode the calculation time.

Beyond just making new cool shapes there has been some - very limited - evidence that having cis peptide bonds in peptide macrocycles may make them more membrane permeable. That is to say, more likely to pass into the blood and then into cells in the body, if we were trying to make a drug from these peptides that could be taken as a pill. Testing this hypothesis has been difficult because before our group (and your computing power) there was no good way to purposefully design peptide macrocycles that contain cis-bonds. Now that we can, I'm making tons of these new designs to see if having the cis bonds does help them cross membranes. If they do that's a great new tool we have when we try to make peptide macrocycle drugs. If they don't, then it stops future researchers from spending time using a faulty hypothesis, and leaves us with a bunch of cool new cis-containing macrocycle shapes to play around with.

I hope that answers your questions. Let me know if you have more.

Thanks. That is a really excellent explanation, from theory to application. It means a lot for what we are doing.

---

Thank you. It's very interesting. My smartphone is crunching...

---

Hi Timo,

The DRH_Curve jobs all belong to me. I am a graduate student in the Baker lab working to design new proteins that can be used as molecular switches. These switches could be used as tools in other labs around the world to control and study the interactions between natural proteins. There are also potential applications of these protein switches to control cell therapeutics such as Car-T cells which are being developed in other labs to teat cancer.

The number of possible protein structures that can be generated on a computer is extremely large and for each of these possible structures, there can be a very large number of possible sequences that we expect to fold to that particular structure. This is where Rosett@home comes in. After we design a particular protein structure with a particular amino acid sequence, we use Rosett@home to check whether that sequence is likely to fold to the desired structure. This allows us to eliminate "bad" designs and do further work with the "good" designs. These good designs are then encoded in a DNA sequence, which is put into a bacterium or yeast to produce the protein in our wet lab, allowing us to characterize the true structure of the protein and see how that matches with our computational models.

Sorry for the high level description of my work which glosses over a lot of background and details. If you have any questions or want more detail, please let me know.

Best,
Derrick

After we design a particular protein structure with a particular amino acid sequence, we use Rosett@home to check whether that sequence is likely to fold to the desired structure. This allows us to eliminate "bad" designs and do further work with the "good" designs. These good designs are then encoded in a DNA sequence, which is put into a bacterium or yeast to produce the protein in our wet lab, allowing us to characterize the true structure of the protein and see how that matches with our computational models.

That is the key point that allows us to see what good Rosetta does. A high-level description is good enough for most of us. We can carry on from there.

---