How Gamers Can Help Fight Covid-19 - Article
by Karl Koebke , posted on 28 April 2020 / 3,011 Views
According to the worldometers website, as I write this article, there have been over 3 million cases of Covid-19 confirmed across the world resulting in over 200,000 deaths. With a lot of our readers stuck at home or working essential jobs it is easy to feel helpless. Staying at home and maintaining social distancing is helpful in and of itself, but it can feel like more of mitigation than a solution. For this article, I wanted to go through a couple of programs that non-scientists can use to help with the efforts against covid-19, either passively or more actively. Below is not an exhaustive list, and while I work in the realm of biochemistry I am by no means an expert on computational design or viruses and their treatment, so you’ll have to forgive me if my explanations are imperfect.
HEWMEN Cell
An e-mail about this program I got this morning was the inspiration for this article. HEWMEN Cell uses your computer’s excess processing power (according to settings you define) to search for drugs against key Covid-19 protein targets. As computers have become more and more powerful they have become an important strategy for drug design. The goal is to find a small molecule drug (i.e. hydroxychloroquine) which can bind to a particular position on a protein of interest and in some way inhibit its function.
Example proteins one might want to inhibit include the Covid-19 protease and viral spike. Proteases are enzymes that cut other proteins at peptide bonds. Sometimes proteins are expressed by viruses in non-active forms and must be cut to their active form, so inhibition of the virus’ protease is a standard tactic to slow its spread throughout a host. The viral spike, on the other hand, functions in how the virus enters a cell for infection, again a prime target for therapeutics.
Because a small molecule drug is tiny compared to the protein being targeted, computational modeling must attempt a variety of positions and orientations for the drug along the protein’s surface. For each position the program will find the optimal binding orientation of the drug and calculate how well the drug binds in whatever pocket it found. While each of these simulations are relatively simple for a given drug/protein pair (using only minimal chemical information), you often set up 100s of simulations to get the best estimate of how the drug will bind. Put that together with a relatively large drug library with 150,000 possible drugs and you can see how this would take quite a bit of processing power.
What Does the Program Do?
HEWMEN Cell takes the known protein structures of about 20 Covid-19 proteins and searches for one out of 5,600 already approved drugs that can bind to some position on the protein. The company then plans to distribute any hits to the relevant agencies so these drugs can be tested for efficacy. Starting with approved drugs is a great way to avoid unforeseen side effects as these are already used in the clinic.
They will also be doing this search with a larger library of 156,000 drug-like molecules to find novel drugs that have not previously been tested in the clinic and would require more rigorous testing before application. Below is a video showing 1,900 drug targets found by a different program to illustrate the kind of output one might get from a single protein structure with a large library of possible small molecule drugs.
What Do I Do?
All you need to do is install the program and set the parameters for when you want it to run. These include things like how many processor cores to use, maximum CPU percentage that the program can use, and thresholds over which you want the program to pause. This simplicity is a benefit for those that want a more passive experience, but the visuals that come with the program are too basic and I would’ve appreciated something more interesting to look at like showing me the drug that the program is currently working on.
Folding@home
Folding@home was initially released in 2000 and has been adapted to a variety of platforms. Personally, I remember installing this on my PS3. Folding@home has a variety of projects for Covid-19 from drug discovery like HEWMEN Cell to folding proteins from their primary structure. For the purpose of this background we’ll focus on the simulation of protein dynamics.
It is hard to define how important structural determinations by x-ray diffraction have been for biochemistry. Under certain conditions proteins are grown into ordered crystals. It is a painstaking process (I can speak from personal experience on this one), but if you succeed you can obtain a picture of what the protein looks like. Proteins are huge and complex, so determining their structure is only possible through a few methods and no other method can give the kind of resolution that x-ray diffraction can.
The above is great, but there’s an issue. These crystals are just a single picture of what the structure looks like, but proteins aren’t static in solution. Rather, they flex and jiggle. This movement is called protein dynamics, and it can be important for drug discovery. Drugs are not targeting proteins in crystal form, but rather how they are in solution, so if the drug can bind to one of these flexed structures it can still inhibit its activity and be an effective therapeutic. By including these movements we can increase the number of target structures for drug design and use structures that are more relevant to how the protein is in solution.
What Does the Program Do?
Folding@home uses molecular dynamics simulations to determine the dynamics of a protein in solution. Starting from a published static crystal structure the program will treat the protein as a bunch of balls attached by springs (atoms attached by bonds) and simulate what happens over a given amount of time. Proteins move quickly, so we’re talking quite a bit of computer time to get a simulation of protein dynamics over a time frame of a couple of nano seconds (1 x 10-9 seconds). Below is a simulation that the team published on YouTube of the SARS-Cov-2 viral spike.
What Do I Do?
This is again a passive way that your computer can do work using excess processing power. You can set some simple parameters like what project you want to work on (Covid-19 or Ebola, for instance) and if you want it working only when the computer is idle, but I’m not a fan of the settings for how hard it’ll work your computer. I could only find three settings: Light, Medium, and Full. I would’ve appreciated more tinkering, similar to what HEWMEN Cell enables. That said, if you care to look you can see exactly what structure your computer is working on, which is a benefit in comparison to the visuals of HEWMEN Cell.
Rosetta@home
Rosetta is a protein folding and design program that has obtained a position of prominence within the field. If you ever want to see what a rock star professor’s lab looks like I suggest checking out the creator of Rosetta, David Baker. With 34 graduate students, 37 postdocs, and a plethora of other members in various positions, this is a lab that I estimate takes 3-4 million dollars a year to keep running just in salaries. The reason he’s trusted with these kinds of resources is because Rosetta helps solve one of the most important problems in biochemistry - how a protein folds into its three-dimensional structure.
The crystal structures we talked about above are great, but what do you do when the people in the lab have not solved the structure for your particular protein yet? Well, technically, if we’re really smart we don’t need the crystal structure. Using DNA and protein sequencing we can easily determine the primary structure (sequence of amino acids) of any protein, and then it is just an issue of figuring out how it will fold.
Proteins are composed of 20 amino acids (with the odd additional one thrown in every now and then), like an alphabet that can be put into an almost unimaginable number of sequences to create the majority of protein structures found in nature. This long strand of amino acids will then fold itself (often spontaneously) into the correct three-dimensional structure. So, if we understand all the rules to how this sequence of amino acids dictates the eventual three-dimensional structure, we can estimate the structure of a particular protein before experimentalists have been able to grow and diffract crystals (a tedious and expensive process).
What makes this difficult is there are a lot of rules that we have determined empirically for how amino acids want to fold up and when you try to combine all of these rules with the often 200 or 300 amino acid long protein the number of possible folding patterns becomes intractable. Rosetta gets around this by first separating the protein into smaller sequences to which it can easily assign a local structure and using empirically determined rules that it applies to the eventual fully folded structure.
What Does the Program Do?
Rosetta@home takes protein sequences from different labs around the world and predicts what their three-dimensional structure will be. When it comes to Covid-19 this could entail viral proteins which researchers haven’t gotten crystal structures for yet, proteins that the virus targets in the host, or proteins that researchers are designing to interact with the virus as a form of therapeutic.
The reason this needs a large pool of CPU time is that the program does not just try to solve the structure once. One of the best ways for programs like Rosetta to figure out a complex problem is to put in an element of randomness and attempt the solution hundreds or thousands of times to find the answers that are either the best (using some scoring protocol) or most frequent. For example, the Rosetta@home project I uploaded said it was going to take a day and a half to complete, though to be fair that’s on a pretty unimpressive laptop.
What Do I Do?
This is the last passive program that I am including in this article. Like the programs before you put in some settings for when you want it to run and how much of your computer’s resources you want it to use. My favorite aspect of Rosetta@home is that it comes with a visually exciting screensaver that visualizes the program solving the protein’s structure. I have included a video below showing someone running Rosetta@home to give you an idea of the visuals. However, Rosetta@home is more difficult to install and get started because it runs through a secondary program called BOINC, which you then need to connect to the specific project you want to work on (in this case Rosetta@home).
Foldit
Foldit is one of those concepts that is brilliant but seems so simple after someone has already thought of it. It takes the problem of protein design, sets up rules and a scoring system, and lets the power of the cloud go to work. For Covid-19 Foldit users are focused on designing proteins that will interact with different portions of the virus to inhibit its function.
The problem of protein design is another layer of difficulty compared to that of protein folding. If the 20 amino acids are like an alphabet, then the problem of protein folding would be trying to read a sentence and interpret its meaning. Basic grammatical rules and syntax allow you to interpret each word from its letters, and the overall meaning of the sentence from the words. Protein design is instead how you write a sentence to convey your own meaning. What sequence of amino acids can I use to create a protein of my own design for a particular function? This is a problem close to my heart as it is the focus of my own research interests.
As I mentioned before there are general rules to how amino acids fold into larger structures that we can incorporate into our own designs. Some amino acids prefer to be on the outside of a protein so they can interact with water (hydrophilic), others want to bury inside the protein to avoid water as much as possible (hydrophobic). You also want to efficiently use space within a protein; there should not be large holes where no amino acid has taken up the space, but you also can’t have them so close together that they clash. Combining these rules together, Foldit’s users can design their own proteins or try to determine the three-dimensional structure of novel proteins.
What Does the Program Do?
Foldit’s genius is that it visualizes all of these rules for how amino acids want to fold in very simple ways like making hydrophilic amino acids blue and hydrophobic ones orange. It then lets users play around with the structure and the amino acid sequence to try and find solutions that a computer may not be able to reach. For Covid-19, users are given challenges like designing a protein which can bind to the viral spike. Promising candidates are expressed and tested.
What Do I Do?
This is a game you can play that might result in a therapeutic for Covid-19. It is the most active role one can take other than being in a lab already studying the problem. You will have to go through tutorials to learn the controls and how all the rules work, but I have included a video below that shows how to get through those as quickly as possible and gives an idea of the visual presentation.
And that’s it. I hope this was informative and helpful. I understand that none of these will solve the problem themselves, and it’s unlikely any single user will be the catalyst that solves this problem. What I’ve learned during my time as a scientist is that it is easy to feel like you are just spinning your wheels working on something nobody actually cares about and won’t amount to anything. Rather than a series of pointlessly spinning wheels, I like to think of it as millions of cogs within the massive machine of science constantly pushing us forward. There are some larger cogs (like David Baker for instance) that will put in a disproportionate amount of work, but we all help. Every little thing we do to help is not meaningless.
As we face down what will likely be the greatest crisis of many of our lifetimes I hope we can keep this philosophy in mind. It’s easy to see these quarantines as selfish, shutting yourself out from the greater world to keep yourself safe. Instead, I think it’s an act of community. People around the world are suffering so that we can protect those who are most vulnerable to this disease. There are reasonable arguments to be made about what strategies will truly reduce suffering to the greatest degree, but the intent of this action is something I will remember forever.
Background links:
https://www.worldometers.info/coronavirus/
https://www.bakerlab.org/index.php/members/
https://boinc.bakerlab.org/rosetta/
More Articles
I'm doing Rosetta for a couple years now already and of course pushed the program in BOINC after Covid-19 to do my part.
After I wrote this I ended up putting rosetta on my own computer and my son's. I like it's presentation the best. Didn't check whether I could set something up on the PS4 like I used to have on the PS3, that would be interesting.
