With the recent COVID-19 outbreak, R@h has been used to predict the structure of proteins important to the disease as well as to produce new, stable mini-proteins to be used as potential therapeutics and diagnostics, like the one displayed above which is bound to part of the SARS-CoV-2 spike protein.

To help our research, we are happy to announce a new application update, and thanks to the help from the Arm development community, including Rex St. John, Dmitry Moskalchuk, David Tischler, Lloyd Watts, and Sahaj Sarup, we are excited to also include the Linux-ARM platform. With this update we will continue to make protein binders to SARS-CoV-2 and related targets using the latest Rosetta source.

Thank you R@h volunteers for your continued support to this project. Your CPU hours are used not only to accurately model the structures of important proteins, but to design new ones as well. Let's band together and fight COVID-19!

More details will be available in the Discussion of this news post.

---

The primary way proteins interact with each other is by sticking to one another. As you may have seen from the R@h graphics application, proteins come in all shapes and sizes. For this reason, most proteins do not stick randomly to each other, but rather stick very specifically to a handful of other proteins. For instance, the viral spike protein of SARS-CoV-2 sticks to the human ACE2 protein which is how the virus gains entry to the cell.

The IPD has been working hard at improving the ability to design such binding interactions. This process starts by creating a set of scaffold proteins that do not have a purpose other than to fold precisely to an atomic structure. These scaffolds are then docked onto a target protein of interest and their surfaces designed to perfectly complement the target. Finally, the designs are scored, filtered, and tested for binding in the lab.

We will now be using R@h to do the surface design step. Docking and filtering are fast, but actually doing protein design is slow. We will be using the massive amounts of compute power available on R@h to sample every amino acid at every position at the interface. We will then pick the best combinations of amino acids using simulated annealing and Monte-Carlo. Sampling is key for this process and this is why we turn to R@h.

So, join us in the coming weeks as we make binders to SARS-CoV-2 and related proteins. We'll still be doing structure prediction and scaffold design as these are absolutely critical to protein science as well. But look out for the interface design cases, because someone might be designing the next COVID-19 cure.

And hopefully you'll stick around once the pandemic is over. We can only design binders like this because we've been working hard at the problem for years. There's still a long way to go, though. Improving the science takes time and computing, so we hope you'll join us on this exciting ride.

-- Brian Coventry

---

A big thank you to Rosetta@home for the inestimable bounty of being able to assist, even if only in such a simple thing as sharing my computer cycles. I am aware that Rosetta@home is one of a number of organisations and groups working to resolve both this current crisis, and the multiple other debilitating condition which affect humankind. To each and every one, kudo's and bouqets.

If the world stands together, then together it will overcome. If the world stands as singles, then each will fall and fail singly.

Romane

Congrats on the update guys. Crunching!

Grazie!

Perfect

What an awesom news! I'm proud, that (even in a really little way) we are helping you scientist fighting this battle. Thanks

Congrats on the update guys. Crunching!

---

I would like to help with my 4 ARM64 and also 6 ARM32 linux mini-boards. But unfortunately the minimum memory they need is near 2Gb and all of mine are modest with only 1Gb. Please consider to compile for ARM32 and also try to reduce the target memory of the programs.
Anyway, thanks for your effort

---

Many thanks!

---

For a better world 🌍

Dear Brian Coventry et alia,
1. Your custom protein binding the Spike protein looks like it will put neutralizing IgGs to shame!!! Very impressive!!! {As an aside each dose of this protein therapeutic will have to be injected as it could never survive a transit of the stomach and upper GI tract with the low pH and proteolytic enzymes that would reduce it to amino acids.}

2. I'm laboring under the assumption that Rosetta code incorrectly programmed the use of the L3 cache and will bottleneck if too many Rosetta WUs are running simultaneously. I limited my use to one WU per 5 MB of L3 cache. Exceeding that limitation slows the entire CPU over 60%. This means that my fleet of Xeon E5s are running Rosetta at less than 20% capacity. Has this been fixed and is it safe to run full force now???

3. I believe that your statement implies that all WUs are contributing to Covid-19 research regardless of whether you include "Covid-19" in the WU name and that minirosetta is doing Covid-19 work as well. Is that what you mean???

4. I don't know if I'm running ARM64 Rosetta aps or not. I see from your Applications page there are numerous versions but when I look at the Properties of a given WU I'm running it says the version 3.76, 4.07, 4.08 or 4.12. What are the Linux-ARM platforms???

5. I suggest you tether these blocking proteins together like an IgM. This would more readily mark them for destruction. Also, it appears that your blocking protein is much larger than an IgG so when it binds a Spike it also blocks neighboring Spikes (like an umbrella) from binding ACE2 and being invaginated.

{Please use black text on white background for your forum to maximize contrast. This faded gray text really strains my old eyes so I never read your forums. I would not have seen this announcement except that you were nice enough to send out a BOINC Notice. Thank you in advance :-}

---

{Please use black text on white background for your forum to maximize contrast. This faded on gray text really strains my old eyes so I never read your forums. I would not have seen this announcement except that you were nice enough to send out a BOINC Notice. Thank you in advance :-}

I can't even touch the rest of your comment - but you can add an extension to your browser to automatically adjust the text - via the Chrome or Firefox extension stores, for free.

It could certainly help your eyes - regardless of which webpages you go to.

---

Thank you, lets all join this effort.

Dear Brian Coventry et alia,

2. I'm laboring under the assumption that Rosetta code incorrectly programmed the use of the L3 cache and will bottleneck if too many Rosetta WUs are running simultaneously. I limited my use to one WU per 5 MB of L3 cache. Exceeding that limitation slows the entire CPU over 60%. This means that my fleet of Xeon E5s are running Rosetta at less than 20% capacity. Has this been fixed and is it safe to run full force now???

3. I believe that your statement implies that all WUs are contributing to Covid-19 research regardless of whether you include "Covid-19" in the WU name and that minirosetta is doing Covid-19 work as well. Is that what you mean???

4. I don't know if I'm running ARM64 Rosetta aps or not. I see from your Applications page there are numerous versions but when I look at the Properties of a given WU I'm running it says the version 3.76, 4.07, 4.08 or 4.12. What are the Linux-ARM platforms???

running these on my ryzen 3800x (8 core 16 threads) does not seem to matter if i run 8 threads or 14-15 threads get roughly same folding times (1h to 30 minutes longer per work unit when at 14-15 threads) are your running them in a hypervisor maybe it's causing problems,
Just spool a xeon system for Rosetta to 50% and maybe another system to 90% to see what happens (just make sure you're comparing similar batch of work units over 3 days)

some work units that are not covid-19 (don't think it is) seem to take long time like this one https://boinc.bakerlab.org/rosetta/result.php?resultid=1138029505 but it did a lot of structures 144 vs the 37 ish in the covid-19 work units https://boinc.bakerlab.org/rosetta/result.php?resultid=1138984983

ARM64 is for ARM64 CPUs its not related to X86 CPUS (AMD/Intel) you can't get a ARM work unit on a normal CPU

covid-19 work units are higher priority at the moment you can still get non covid-19 work units at times (as compared to last year there are now 8x more CPU power available then september last year so they are been worked on very fast, so could be times when your getting non covid19 work units)

the covid-19 norm has covid-19 in work unit name, did look like they ran out of work units at one point as it was doing very long 18 hour work units that i not seen before,but its working on new batch of covid-19 now (haivRoid vs covid19)

Also I say hi as this is my first post on here (the android client needs some love as it's not very reliable at not destroying work units that are on the phone, as I come back and they are gone lots of download errors on website and pending work units that no longer on my phone)

Will you make any COVID-19 badges ? ;-)

---

Great news Rosetta, seems there is an app for android and an app for Linux on ARM64bit, assuming my rock64 will run it, not sure about my Pi3 & Pi2 tho...

Will you make any COVID-19 badges ? ;-)

That would be nice.

---

we are excited to also include the Linux-ARM platform. With this update we will continue to make protein binders to COVID-19 and related targets using the latest Rosetta source.

This is really exciting! Is it sufficient to run boinc on an ARM machine and it will start crunching R@H units? Does this also include Boinc on Android?

Michael

How can I join my computer cycles to the covid19 project?

[snips]

2. I'm laboring under the assumption that Rosetta code incorrectly programmed the use of the L3 cache and will bottleneck if too many Rosetta WUs are running simultaneously. I limited my use to one WU per 5 MB of L3 cache. Exceeding that limitation slows the entire CPU over 60%. This means that my fleet of Xeon E5s are running Rosetta at less than 20% capacity. Has this been fixed and is it safe to run full force now???

3. I believe that your statement implies that all WUs are contributing to Covid-19 research regardless of whether you include "Covid-19" in the WU name and that minirosetta is doing Covid-19 work as well. Is that what you mean???

running these on my ryzen 3800x (8 core 16 threads) does not seem to matter if i run 8 threads or 14-15 threads get roughly same folding times (1h to 30 minutes longer per work unit when at 14-15 threads) are your running them in a hypervisor maybe it's causing problems,
Just spool a xeon system for Rosetta to 50% and maybe another system to 90% to see what happens (just make sure you're comparing similar batch of work units over 3 days)

some work units that are not covid-19 (don't think it is) seem to take long time like this one https://boinc.bakerlab.org/rosetta/result.php?resultid=1138029505 but it did a lot of structures 144 vs the 37 ish in the covid-19 work units https://boinc.bakerlab.org/rosetta/result.php?resultid=1138984983

covid-19 work units are higher priority at the moment you can still get non covid-19 work units at times (as compared to last year there are now 8x more CPU power available then september last year so they are been worked on very fast, so could be times when your getting non covid19 work units)

the covid-19 norm has covid-19 in work unit name, did look like they ran out of work units at one point as it was doing very long 18 hour work units that i not seen before,but its working on new batch of covid-19 now (haivRoid vs covid19)

I'm not sure the use of the L3 cache CAN be programmed, unless you are working for Microsoft and are writing code that will be part of Windows. The path from the CPU to memory outside the CPU is more likely to be overloaded for CPUs with large numbers of cores, and therefore slow down all tasks for that CPU.

The first type of tasks for COVID-19 had COVID-19 in their names, but it is no longer the only type. Some of the more recent types had foldit or design in their names instead.

Also, tasks submitted through the Robetta interface have no information in their name about whether they are related to COVID-19 or not; these have names starting with rb_.

Having an SSD drive on your computer and running all BOINC work from that drive helps with disk access speed.