Big Compute - The Clever Coatings of Coronavirus

Episode Date: October 20, 2020

It’s been months since the infamous coronavirus has crept across the globe, closing schools and workplaces and changing the way we live our lives.  But why is COVID-19 seemingly so good at infecting people?  What makes this virus different than others?  We talk to undercover superhero, Rommie Amaro of the University of California San Diego, about her discoveries through computational simulation of what the virus actually looks like, how it moves, and what that means for each of us.

If I saw Marshmallow Peep Sugar doing that, it would freak me out. You would eat it. Hello, everyone. I'm Jolie Hales. And I'm Ernest DeLeon. And welcome to the Big Compute Podcast. Here we celebrate innovation in a world of virtually unlimited compute. And we do it one important story at a time.

We're talking about the stories behind scientists and engineers who are embracing the power of high performance computing to better the lives of all of us. From the products we use every day to the technology of tomorrow, high-performance computing plays a direct role in making it all happen, whether people know it or not. Yes. So on our last episode, we spoke about computational simulations that were run to see how COVID-19 spreads through the air indoors. And we talked about which musical instruments are at higher risk of spreading the disease. Which included giving props to one of my favorite instruments, the sousaphone. We didn't talk about the sousaphone.

It's just a wearable tuba. That's all it is. Clearly I'm not a tuba player. Okay, the sousaphone is a wearable tuba. Wow, I learned something. And I minored in music, so that was a fail in my education. Any marching band you've ever seen? You know, the tubas that they kind of come up and face forward?

Yeah, the ones that are all wrapped around them like a snake? That's a sousaphone, yeah. But it's literally the exact same as a tuba other than the way you wear it? Yep, same key, same everything. What? Named after John Philip Sousa, one of the most famous wear it? Yep. Same key, same everything. What? Named after John Philip Sousa, one of the most famous composers of all time. Yes, that makes sense.

Huh. Thanks for teaching me something. Today, we're going to continue down a similar path, not so much with musical instruments, but with... The coronavirus. Coronavirus. The coronavirus pandemic, because new cases are on the rise. similar path, not so much with musical instruments, but with the coronavirus, coronavirus, coronavirus pandemic, because new cases are on the rise. The new numbers not seen since the summer,

which is unfortunately still very much on the forefront of our lives. It's affecting our health, our schools, our places of worship, our workplaces. It's really hard to find something that hasn't been affected. I mean, as just an example, I worked for the Walt Disney Company for about eight years. Please welcome our Disney ambassador, Jolie Hales. I represent the 23,000 cast members who work at the Disneyland Resort. Maybe 23,001, I don't know. That extra one adds a lot, yeah. And I built a lot of really great relationships there because there's a lot of amazing people who work for Disney. And then recently, over a period of just a few hours, I literally watched on my Facebook feed as hundreds of my Disney friends lost their jobs.

But we start first with breaking news out of Disney. Disney is laying off 28,000 domestic employees at its parks experiences and product segment. And that will affect employees across all levels, hourly, salaried and executive roles. For some, it was the only job they ever knew. And it was this weird feeling to have this sadness kind of hanging over the happiest place on earth. And I know that other industries like airlines and theaters are seeing similar losses. That's true.

As a matter of fact, today I was reading an article that Cisco is doing some massive layoffs right now. And the sentiment is that it's probably the largest layoff they've ever had. Oh, man. The thing about this virus or a virus in general is that it doesn't care what your job is. It doesn't care what your wealth or social status is. It's perfectly fine to invade your life regardless. Just look at the President of the United States. Yeah, people were kind of flipping out over that one. Wasn't feeling so well. I feel much better now. We're working hard to get me all the way back. So today we're going to talk about an undercover superhero who is working with high-performance computing to do something that we are all rooting for.

Find a cure for this stupid virus, or at least understand it enough to get to a vaccine and better therapeutics so that it's not so deadly and not so interruptive. I think at this point, we're all ready to just heal and get on with our lives. So true. So, if you search the interwebs for COVID-19 research, the name Romy Amaro pops up everywhere. And for good reason. Romy is...

I'm a professor of chemistry and biochemistry at the University of California, San Diego. Where she works in? What we call computational chemistry or computational biophysics. It's basically that we are using mathematics and computing to understand better sort of biological and chemical systems. And not only is Romy a kick-butt scientist, but she's just an all-around cool human. I have four children. Awesome. So I spend a lot of time,

you know, right now doing sort of the remote schooling business that we're all, you know, many of us are sort of dealing with. I mean, one of the good things about having so many kids is like, we're not necessarily really lonely here. We have a lot going on. So that's a good thing. It's always a party at the Amaro home. Exactly. Exactly. Exactly. Birthdays come and we can still have a crowd for singing. So that's good. And when she can break away. But I'm also a runner.

I've been a runner since I was really young and I enjoy the heck out of that. So I try to do that, you know, as often as I can. Uh-oh. There was an audible gasp there. I heard how excited you got to talk about running. Yeah, that got me a little off track for a bit. What's your favorite distance to run? You know, I think the best distance is a half marathon.

That's my favorite too! Oh my gosh. But after some solid blissful moments discussing the joys of repetitively putting one foot in front of the other until you almost collapse, Romy told me that during the times that she's not running or with the family, she splits her work time between teaching and doing research. I have to say we do spend quite a bit of time doing the research part of things. I'd say that's understandable

given the state of things right now. Everyone wants to know how to put this pandemic to bed. My group jumped into studying COVID-19 really with all feet and it's been so intense. And when she started this research, she began to see something in the scientific community that she had never really seen before. By mid-March, we had already mobilized and said, hey, folks, we all need to work together on this like we've never worked together before. Because typically, you know, academic researchers are kind of weird. Well, you know

that you're like, yeah, tell me something I don't know. But we tend to study things and be very possessive about them. We don't share with people necessarily unless we really trust them or unless they're collaborators. And we we certainly wouldn't share something very early. Like you get a little bit nervous about sharing things because, hey, you might have a mistake in there. That's part of it. People are insecure. But also there's a competitive edge to it, right? Where like you want to be first because in science, it's a lot about being first and making that first discovery. So there isn't really a sense of sharing generally to the extent that probably we should. But that was completely not the case with COVID-19. And so we just said as a community, actually, we had over 200 different

groups come together and say, we're going to share things as quickly as we can. We're going to build these models together. Yeah, so it's been really cool. So we have had to stay really organized because we actually have shared this data now with people worldwide. Yeah, that's absolutely true. I was even surprised to see how this played out in terms of research and science. The scientist community decided that it wasn't important to find out who started this or why. What was important was to figure out what was causing this and how could we solve it. So in this case, it was amazing to watch these scientists just come together globally, ignore all of the white noise, and just get to work on trying to find a solution for humanity.

Yes, this kind of global collaboration in the scientific community, as you said, is kind of unprecedented. I mean, if you think about it, I don't know, maybe what makes the COVID situation different is that every scientist and researcher out there has family members and friends who are in that high risk category for the virus. It pretty much hits close to home for everyone. And researchers know that it's kind of on them to basically fight for the human race. So how did Romy and her colleagues first get involved in this research? Were they working on something else and they dropped it to study COVID? Or were they looking for something else to study in particular?

Well, the timing is pretty interesting. Romy and her team had actually just spent five or six years studying influenza. You know, that common flu that we've all had and that we all completely hate. When you say we've all had, you must mean the royal we. I've never had the flu or a flu shot, but. What? You've had the flu. Never in my life have I had the flu or had to have a flu shot.

What? You're an anomaly, Ernest. Nobody who's listening to this can relate to you. I'm telling you. I get the flu every year and I get a flu shot. So I don't know what's going on with me and my immune system. We need to study you is what we need to do. Yeah, I'll donate my body to science once I've exited this world.

Which hopefully won't be anytime soon. That's the hope. How do you distinguish between coronavirus and the typical flu? It can be a challenge when a patient first presents for care because the symptoms overlap quite a bit. There was some news about it already kind of trickling around and my colleagues and I started just kind of watching what was happening in terms of spread. So they spent all these years putting together a study about influenza that was actually published in February, right when COVID was hitting the news in like a massive way.

Because at this point in early February, it had really started to take off in Italy. The prime minister has put a total lockdown for all 60 million people until next month. Italy has the biggest cluster of cases outside of China. And that was when we said, OK, I think I think we really need to sort of pivot our efforts and try to see what we can do with this. And then after that, it came very quickly to the United States and things really took off. It sounds like they didn't have a lot of time to celebrate the completion of their influenza study. They had to pivot nearly immediately from one study to another. Yes. So if they were expecting a break, they didn't get one. So what about COVID did they study?

Okay, I'll tell you. But before I do, I want to quickly review how viruses work. Enlighten us. So, in a nutshell, viruses are infectious microorganisms that need a living host to survive or to multiply. So, they're obviously too small to see with the naked eye. In fact, according to Dr. Clayton Cowell of the Mayo Clinic, the COVID-19 virus is 1 900th the width of a piece of hair. 1 900th? Yes. So as we're just going about our business out there in the world,

if this tiny virus is able to get into our body, either through an infected person sneezing or by licking the infected surface of a telephone pole for some reason. Or, as we're finding out with COVID, just standing indoors near an infected person who is talking or even just breathing. Or an asymptomatic person playing the trumpet in the same room. But not the tuba. Nope, tuba's good. So, in any of these infectious situations, a virus can enter our body through the mouth or nose or whatever. And then it tries to attach itself to one of our living cells because that's how it basically survives. It needs that living host.

And if it's successful, it does everything in its power to start multiplying and attach to more and more cells, often making us feel like garbage along the way. While I've never had the flu myself, I have had bronchitis several times. And I can tell you, being sick with an upper respiratory infection is the worst. Oh, amen to that. And I'm still absolutely blown away that you've never had the flu. Like seriously, scientists listening to this need to like get blood samples or something and figure out what the deal is because people like

me want a piece of that action. I don't know. I'll take a blood transfusion from you, Ernest, if that's going to cure me of all future flus. I don't think that's how science works. But I mean, it might prevent you from getting a viral infection, but bacteria, they're safe. Gotcha. But I mean, it really is crazy how something one nine hundredth the width of a piece of hair can have such a massive effect on us. But after the virus has infected us, if things go well, our body recognizes that we've been invaded. And then it launches a super epic immune response where this huge army of defender cells comes out of the woodwork, or I guess the blood work in this case, and sets off to seek and destroy all invaders.

So it's like that scene from Lord of the Rings where you're having the Battle of Helm's Deep and it looks like all is lost. And then all of a sudden Gandalf appears at the top of the ridge with the Rohirrim behind him and the sun glaring down behind him onto the army of orcs. Yeah, it could be that. I like to picture it more of like a space invader fight. I picture a Star Wars fight, but I like the lord of the rings take um in fact okay so part of that immune response in our bodies is executed by what are known as t cells and you've

probably heard of them before there are x wings and a wings t wings really doesn't have the same ring to it okay you're right i guess it sounds more like a cartoon bird drinking tea or something but anyway you get the right? These T cells go to work killing off the virus invaders. And then another part of our immune system called the B cells starts making antibodies, which are special proteins that basically have two main jobs. So first, they bind themselves to the virus to stop them from multiplying. And second, they kind of stamp a label on the virus as being an invader so that other cells know to immediately destroy any more that they come across in the future so that's why if the

same virus strand enters your body months later it's typically identified immediately and destroyed before it has a chance to multiply and make you sick again so is the problem that viruses just aren't discovered by our immune system quickly enough so they have enough time to multiply and invade the body before an immune response can actually take over? I think so. I wonder if there are really obvious viruses that our body recognizes and destroys so quickly that we don't even know we were infected. Yeah, like they aren't good at blending in with all the other cells. So the security alarm goes off right when they appear in the body or something. My guess is that it happens maybe with really wussy viruses. But then again, maybe those wussy viruses wouldn't survive enough to be passed to another person at all.

So natural selection might just eliminate them. I don't know. But then there are other viruses that I know are especially deadly because they kill off our immune system before it can win the fight. And this was the case with HIV for a very long time until very recently where we've gotten the advanced therapeutics to actually deal with it. Exactly. Exactly. HIV or makes HIV such a dangerous virus, because after it invades the body, it immediately attacks and kills the T cells, which is that immune system fleet that typically seeks out and destroys the invaders. It's mysterious, it's deadly, and it's baffling medical science. Acquired immune deficiency syndrome. If the trends continue as they are, I think we can predict that the acquired immune deficiency

syndrome is a highly fatal illness likely to remain with us for the next decade. So with all this in mind, let's go back to Romy and what she and her team were trying to learn about coronavirus. We were very interested to study the main infection machinery of the virus. They were particularly focused on the part of the virus that actually latches on to human cells causing an infection. And we've all actually seen images representing that molecular machine, whether we realize it or not. The iconic image is the one that they show where it looks like a gray golf ball with like little red spikes coming out of it.

Yeah, I see that all the time. Right. So that's the virus. And the red spikes that you see in all those images, those are that's the actual spike protein. They actually call it the spike protein. Well, that makes sense. Isn't that cool? Finally, somebody in science that can name something the way I would have named it. There's spikes here. Let's call it a spike protein. That's exactly right. Spoken like a true scientist. Always. And each particular virus, you know, on average might have about 30 to 40 spikes per little gray ball.

Okay. And that's all it has. You know, inside it's got all this other bits that it uses to carry out the infection process. But otherwise it's like, you know, the virus is like a little container that has these spiky bits coming off of it. And it's these spiky bits that ultimately encounter the host cell or come up and sort of touch or hit the host cell. And then that cell essentially can become infected. So Romi and her team were focused in on those spike proteins. Yes. And as a part of that, they wanted to learn more about how or why this particular virus was so good at infecting people, why it was able to latch on to human cells so effectively.

And Romi said that each COVID-19 virus basically has about a dozen different what she calls machines on it that serve different purposes. I'm studying just one of the machines, but I think it's among the most important of the machines because it's the machine that literally will latch on to your human cell and ultimately like infect you. What a jerk machine. I know. It's pretty dangerous. It's very clever. I'm going to tell you what a clever machine this darn thing is, unfortunately for us and for our immune system. So through simulation, Romy and her team zoomed in on this particular machine or this particular part of the virus spike protein.

If we could see what it looks like and if we could understand how it moves, then maybe we could figure out a way to break the machine. And by breaking the machine, then we possibly could stop the virus from infecting human cells. So that was the goal. Figure out how this microscopic bully golf ball with a bunch of spikes on it is latching on to our cells and making us sick. And then use that information to destroy it. Sounds like the work of an undercover superhero. And then get this. She started to explain that apparently in our bodies, our cells are normally covered with this sort of sugary coating. And that admittedly got me really interested because anyone who knows me knows that I'm slightly and unfortunately obsessed with sugar.

For instance, I love marshmallow peeps because they're basically gelatinized sugar covered in crystallized sugar, which is amazing. Probably not the same kind of sugar as the kind coding human cells. Sadly not. But I might just picture our human cells as a bunch of marshmallow peeps anyway. You do you. Oh, we could even combine. Oh, now I feel like I'm like in a children's classroom or something. But what if we combine the space spider visual with the marshmallow peep idea?

Like the immune system is an army of yellow marshmallow peep chicks and space spiders taking out viruses. I think you've ventured into the territory of a child's imagination. Okay, okay. But seriously, going back to the science of it all, the crazy thing that Romy and her team have discovered relating to this sugary coating on our cells is that the virus almost seems to know about it and use it against us. And how does it accomplish that? I'm going to tell you after the break. They call it high performance computing, but is it really living up to its name?

I mean, how much has really changed in the last 15 years? Since the revolutionary jump to cluster computing, there have been new core types, new ways to queue jobs, but no real seismic shift. Until now. Introducing Rescale. Until now. Visit rescale.com slash bcpodcast to learn what a modern approach to HPC can do for you. Rescale. Tomorrow's HPC, today. Okay, where were we? You just told us about the sugary shield on human cells. Yes, the marshmallow peep space fighters.

And you were just about to tell us how COVID fits into this picture. Yes. So through their simulations, Romy said that they likely discovered why COVID-19 is so good at infecting the human body. It turns out that the COVID-19 virus is also actually covered in its own sugary coating. What the virus will do is it basically tries to hide itself from your immune system. And the way that it does this is by cloaking itself in a shield of sugar. And so by sort of covering all of its bad viral bits, I'll call them, then the human immune system doesn't sense that the virus is in your system. Instead, it just sees this sort of sugary coating and says, oh, nothing to worry about. I'm going to, you know, look for other invaders, you know, in your body. That's so crazy. It's like camouflaging itself.

Absolutely. And tricking the immune system. Yes. Man, what a little sneaky snake. I know. That is clever. COVID-19 is actually disguising itself as a marshmallow peep. So the fleet of immune system peeps doesn't suspect the invader.

Yes. So one thing we know about DNA and nature in general is that it is ultra efficient and it goes with what works. It's very rare to find something that is truly unique and not applicable in any other way. So is this sugary coating, but they have it to different extents. This particular spike protein, this is pretty well-coated. It's similar, I think, more to HIV than it is to the flu in that sense, in terms of how much of the surface is sort of hidden by this sugary shield. So that's a very interesting comparison she draws here between the COVID-19 spike protein and the way HIV operates. Yeah, it sounds like there's a lot of different viral particles that do the same thing,

but COVID is, for some reason, especially good at it. I mean, one of the reasons why we became interested in it as opposed to the structural biologist down the street or another scientist is that we do, as I mentioned at the beginning, we do computational chemistry and computational biophysics. And one of the really interesting things about the sugary coating is that they can't really take pictures of it. So experimentalists can't take pictures of it, and they can't really figure out exactly what it looks like. So the only way to do that is using methods like ours, this sort of computational methods to basically predict sort of what the shield looks like in detail.

So we that's one of the reasons we became interested because we realized, hey, there's these people taking these amazing pictures of this little molecular machine. And they are fantastic pictures, but they don't tell us the whole story. And if you don't know the whole story, well, you know, part of the story is good, but we want to know as much as we possibly can about this darn virus, right? Because the more we know, the better we can equip ourselves to potentially fight it. And they're fighting it by taking that detailed information about the holes in the sugary shield that reveal that there isn't actually marshmallow or whatever underneath or chinks in the armor, if you will. Then Romy and her team provide that information to that large global scientific community, including developers of vaccines and therapeutics, is using that very information from this very study to find ways to break through COVID's sugary shield. That's an awesome impact right there.

Right? I mean, imagine on the day when a vaccine is finally generally available. Romy could quite possibly be at the health clinic waiting in line for a vaccine that her research actually helped create. And nobody around her will be the wiser that this woman standing right next to them has had such an impact on the world. That's what we call an undercover superhero. Is it not the perfect description for her? I mean, talk about a selfless field of work to go into. And all of this is made possible through supercomputing. In fact, that's the only way this sugary shield can really be observed,

by creating a detailed visual through simulation. And so, Ernest, I wanted to show you a couple of these images that were computationally produced by Romy's team. And we'll post these also in the show notes as well on bigcompute.org for our listeners who want to take a look at them themselves. So here's the first one. And I want you to describe, since they can't see it, describe to our listeners what it looks like to you. First of all, the entire thing looks like something out of Yoshi's Woolly World. But there's like a base layer across the bottom that has a bunch of different colors on it.

And then what appears, I'm guessing this is a spike protein coming out of it in a kind of a turquoise or teal color. And it looks like a, almost like a tree trunk. And then what would be branches filled with leaves of this, you know, cottony looking, cotton candy, actually, cotton candy looking teal thing. It's interesting to see that spike protein that we've seen on the golf ball zoomed in. This one's blue instead of red. It's actually really beautiful the way that they've simulated it. And then there's the second image and tell us what it looks like in relation to the first image. So the second image is literally the first,

but the difference is it's covered with these blue kind of fuzzball looking things, which I'm assuming are the sugar molecules that are coating the spike protein. Yes, you're exactly right. So it's basically the same image as before, except this one has that sugary coating. Exactly. And it does look like fuzz balls or like the craft store things that you glue on when you're seven years old and you glue it on like your paper plates and give it to your dad for Father's Day or something. It totally does look like that. It does. It's almost like a dark blue cotton ball stuck all over it. Yeah. So it's pretty interesting to look at these images

side by side because first we have a close-up view of a COVID-19 spike protein, though this one, like we said, is blue instead of red. And then we have the same view, but with that sugary coating added to cover the surface of that spike protein. Right. And it's really good to see the visualization here because now you can see what Romy meant when she said that it really covers up the spike protein really well. There are obviously some areas that are still exposed, but by and large, the majority of the surface is covered. And when you say over the surface, it's really over the surface. Barely any of the surface is covered. And when you say over the surface, it's really over the surface. Barely any of the protein can be seen through that sugary shield. So we can see how it's disguising itself to look like any regular sugar-covered human

cell marshmallow peep. And I mean, if I were part of the immune system patrolling the human body for intruders, I probably wouldn't attack it. Maybe this is part of why COVID-19 is so contagious. It's just too cleverly disguised to be detected fast enough. It looks like it. When the human body sees a COVID-19 virus, it basically just thinks it's supposed to be there, and then it doesn't try to destroy it. So then that virus latches onto the human cells with those little machines on the spike proteins, and then it starts to multiply, sometimes making the human host quite sick. But by using supercomputing, we can actually in silico sort of make this model. And then what we do is we use

the computer again, and in a really sort of compute intensive way, we run these things called all-atom molecular dynamic simulations. But basically, we can not only see what the sugar looks like, but we can also understand how it moves. They've even published some short animations that show these simulated movements, though I honestly need a scientific interpreter to tell me what they mean. To me, it just looks like a lot of colorful springs and blobs jiggling around. But for a trained eye, these visuals are extremely valuable. Here, I'll show you, Ernest. Maybe you'll understand them more than I do. Let's see here. Yeah, it's hard for me to

tell since I'm not a virologist or immunologist, but I can definitely see a lot of movement in this simulation showing where things are kind of attaching and how it's attaching in general. So it's a really interesting and cool simulation to look at. And I'm guessing that it's even more interesting if you're a scientist in this field. That's what I'm thinking. One of the reasons why people before us weren't able to see the full spike protein, what it looked like, is because these sugars, you know, when you think about the sugar on the marshmallow peeps that you love to eat, you think about like hard crystals, right? Yeah. But this sugar moves. It moves more like the branches and leaves on a tree. And so it creates this like blurring effect. And that's part of its like cloaking ability or its camouflage is that it's these sugars are kind of like sweeping around the

surface of this molecular machine and thus like making this sort of almost more like a cloud of sugar to protect it from the human immune system. What? That's so crazy. If I saw marshmallow peep sugar doing that, it would freak me out. You would eat it. Yeah. Okay. Let's be honest with ourselves. Did she tell you how she ran the simulations? Yes. We run these things called molecular dynamics simulations, which are these highly detailed physics-based dynamical models of systems.

In this case, it's a biological system. So what we do is we take that spike protein with all the sugars and we embedded it in a viral membrane. And it's this miniature three-dimensional model. And we approximate that model at the atomic level. So I, of course, understood every word of that. But for those out there who might need some clarification, Ernest, do you care to interpret?

Essentially, what she's doing is modeling this at a very, very small, as she said, atomic level. And then she needs to take that and put it into motion. Oh, and I shouldn't say that supercomputing is the only thing involved in research like this. So often it starts with hands on experimentation and then computation actually amplifies the efforts and takes it further. We start with a particular configuration that is specified by experimental data as much as possible. This is our starting condition at t equals zero. And it's all built. It's got all the pieces. And then we do one integration step and we get another structure.

We see how it's moved in time. And this time step is really small. It's like one or two femtoseconds only. What seconds? Yeah. Do your listeners know about femtoseconds? They probably do. Femtoseconds.

Yeah, that was a new one for me. Apparently, it's 10 to the minus 15th of a second or one quadrillionth of a second. That's a one with 15 zeros behind it. So it's a really, really, really small little stretch of time. Buckle up. It's going to get technical. But it has to be a really small stretch of time because if you take if you're if your integration step is too big, then you violate the physics of the system. You basically clash atoms into each other. And when you do that, then you're totally screwed and your system basically blows up.

Oh, jeez, don't want that. Which is not good. No, you don't want that. So basically, we perform this numerical integration on these big supercomputers. And we do this integration this integration millions and billions and even trillions of times. And that's how we build up, taking very, very small steps, we can build up a dynamical trajectory of the system in time.

Sorry, my brain just exploded. So for our listeners who may not be directly involved in this type of science, think of this like a motion picture. You know, before we had movies as we know them today, a motion picture was exactly what it sounded like. It was a series of pictures that were taken in sequence, often very rapidly, and then strung together to make it look like things were moving, even though it's individual frames. This is exactly the same thing, just on a much, much smaller level and a much more contained or constrained set of data. So what machines were they using to run these simulations? Well, they were using a pretty well-known supercomputer, a machine called Frontera, which is at the Texas Advanced Compute Center.

And remember how the scientific community has come together in an unprecedented way to fight COVID-19? Well, the same thing can be said for the high-performance computing industry. Back in February, when we were just really seeing all the cases take off in Italy before it had hit the U.S., there might have been like less than 10 cases here, like early February. At that point, you know, we saw that it was a pretty big protein and it was going to take a lot of compute. So I actually, I sent a quick email to Dan Stanzione, who's the director of TAC, of T-A-C-C. And I said, hey, Dan, you know, I hope you're doing well. Yeah, I don't know. I don't know what you guys are up to. But, you know, this this virus, this looks this virus is looking

pretty serious. I think we probably need to do something with it. And again, this is really early days. And he just said, you know, Romy, I totally hear you and we're going to make this machine available to you to use. Wow. Yeah. It was amazing. I mean, it was amazing to have that level of support. You know, in part, we have trust. We've worked together on other projects before, but that was really key in sort of getting time to solution very quickly for this effort. Anyway, and I was so grateful for them. So the TACC provided the necessary compute resources for Romy's study, and they did it quite quickly.

In this community, this is what happens. Right. So I know that many tech companies also stepped up to join in this fight. You're absolutely right. In fact, I was personally involved in the communications effort behind a program called Tech Against COVID. And it was basically where companies like Microsoft and Google, AWS, and Rescale, which is a sponsor of this podcast, they had joined forces to provide free compute to scientists and researchers who were fighting COVID-19. And in order to get approval from them, it usually takes weeks. But we had four of these kinds of companies working together,

and they literally pushed out the program and the communication about it in just a matter of hours. And I've honestly never seen that before. And Tech Against COVID wasn't the only program out there that did this. I don't know if you've heard of the COVID-19 HPC Consortium. I have heard about them, but I have a feeling you're going to tell us a little bit more about it. I am. I was on their website yesterday for this very reason. So the COVID-19 HPC Consortium is another resource hub that and this one was actually spearheaded by the White House, where high performance computing resources are donated from, again, a ton of different tech companies. So we're talking Google, NVIDIA, Microsoft, IBM, AWS, Hewlett Packard. There's

a whole bunch of them combined with academic supercomputing centers like TAC, as well as various government agencies and the Department of Energy National Laboratories. I mean, it's a really big group, all donating resources. Ordinarily, it takes us a really long time, a pretty long time, like on the order of months to get access to the big machines that you need to run these kind of computations or these simulations. And one of the amazing things about COVID-19 or one of the, I guess, silver linings is just that everybody decided to be much more cooperative and to step forward to support researchers to say, hey, if you have a need for compute, you can come to us and we're going to turn this around in like

a couple of days or a few days and give you as much as we can. So as of this recording, the consortium website says that there are 87 active projects. And I know that the Tech Against COVID program has active projects as well. So it's pretty neat. I know that with Romi, the research that they're doing is already being used across the globe. And in the HPC world, there's a lot of buzzwords we like to use, performance being one of them. But another one we love to talk about is scale. So with Romi's research, what kind of compute are we talking about? Okay, so for the spike protein study that we've mentioned here, they use 256 nodes per run with 56 cores per node, which if you do the math ends up being 14,336 cores per run. And if you

take a look at Frontera, which is considered to be one of the top 10 supercomputers in the world, it has 8,008 available compute nodes. So just to give you an idea of scope, it would have taken up around 3% of the entire Frontera machine at one time just to do a run to get this look at the spike protein sugars. We were able to utilize, you know, a rather large chunk rather quickly because some of our runs we can actually sort of set up the different systems basically in parallel to each other. So we're using multiple 256 node chunks at a time. And just to bring that home for someone who might be less familiar with supercomputing specs. is parallelized in a certain way, but it would be probably over 700 years of equivalent of a single

desktop. And I would say that's probably what we use up in about two months or so. So from 700 years down to two months, I'd say that's a bit of a difference. And what software do they use? So we use this amazing code called NAMD, which is a molecular dynamics engine, nanoscale molecular dynamics engine, or NAMD. It's developed out of the University of Illinois at Urbana-Champaign, where I got my PhD. Oh, nice. Yeah, yeah. It's cool. Connection. And not only were they using NAMD software, but they got to help kind of customize it as they went, which is unique. We work together not only with the team at TAC on the hardware, but then, of course, also we work really closely with the software developers

like the NAMD team. There's Jim Phillips and John Stone and David Hardy. There's a whole bunch of folks who are working to tune and optimize the code to these particular architectures. And that also makes a big difference. Just trying to squeak a couple more flaps per run gives us a great advantage. So we've worked really well, I think, as a team all together in those aspects. If there's one thing you know about the software development world, it is that as time has gone on

and compute power has increased, there's been a general movement away from software efficiency. So it's amazing to watch or hear that Romy was able to work directly with the writers of the software to actually improve its performance on the specific supercomputer she was using so that she could eke out just a tiny bit more performance to make the simulation run a little bit faster.

And to go a little deeper for you, Ernest, the code is in C++ and then it uses this thing called Charm++ for the parallelization. And then most of our work, we sort of interface through TCL or through TCL interfaces. Now that's not too surprising to hear because again, in the software development world, we use a lot of interpreted languages nowadays for a bunch of different things. But when we really need a heavy hitter in terms of performance, we go to compiled languages and specifically we want to get really low. So we go into C++. And as they're doing all of this, they thought, why stop at just

looking at the spike protein when there's obviously so much more of this virus, this golf ball with spikes on it that can be explored through computational simulation. The other thing that we're working on, which we haven't published yet, but we're really working hard on, is actually moving beyond studying the single spike to actually simulating the whole virus. And so, you know, that takes our problem up an order of magnitude or more. And so there we do have some prototype systems up and running now on Frontera of the full virus. And there we're using 2048 nodes. So, you know, over 2000 nodes of the machine. Wait, 2048 nodes of an 8000 node machine. If they're running that all in parallel, we're

talking about over a quarter of the machine. All for a virus that is one nine hundredth the width of a piece of hair. We are learning things that really you cannot see in other ways. There really is no alternate experiment that could be performed that would give you what we can provide with these simulations. It is really unique. And that's one of the reasons why it's so cool and has kind of made a splash is because, you know, we showed for the first time what it looks like. People hadn't seen that before and they couldn't without compute. So simulations in that sense have a very unique role to play in this sort of space of research in many different areas, but especially in the

case of COVID-19. And while computing is doing a lot, there is still opportunity on the horizon to do even more. There are still sort of gaps in our capabilities in terms of being able to simulate, for example, long timescales or for very complex scenes that have like many molecular piece parts. I think combining these types of simulations together with artificial intelligence, new methods coming in artificial intelligence, that's something that really is just starting to sparkle on the horizon. That's amazing. And artificial intelligence is one of the areas that I really

love to be involved in, think about, especially in the HPC context. And it's for reasons like this. Artificial intelligence is what's going to allow us to shift from kind of what we know and what we do and looking in on something that has already existed or currently exists to predictive analytics, being able to predict something that will happen or predict a movement of something. I'm one of probably many people who really believe that the real breakthroughs in research are going to come at the intersection of simulation sciences with experimental observational science. That's sort of what I was getting at in the last sort of example with AI. But like when we can combine what we can measure with accurate models that can predict forward, then we can get ahead of a lot of these problems, you know, and we could also maybe have a chance to do things like try to thwart climate change or to develop better sustainable energy methods and materials, you know, or develop new therapeutics. I think that's the direction that I'm most excited about.

So with everything Romy and her team have learned so far, I asked her what advice she would give her friends and her family or what she tells her friends and family when it comes to COVID-19. Wear a mask. Oh, yeah, for real. Because really, that's one of the things. I mean, it ties into that because it is a very clever virus. And it has a lot of ways of being clever, a lot of ways that we don't understand. But this just goes to show you don't want to catch this. And it's like, to me, when I first saw this, it was like, oh, my God, this is a very smart virus. And it's even yet another reason to wear a mask and just to be so careful and just to continue to be vigilant. We feel like we've been at this

a long time right March 13th was a long time ago it's like 200 some days everybody's getting burned out but the thing is the virus doesn't get fatigued we still do not have immunity it is still super damn smart with a great camouflage mechanism so i don't know i mean it's tough but we gotta stay the course stay the course earnest or risk getting sick via marshmallow peep imitation i feel like i'm going to dramatically regret the marshmallow peep part of this whole episode later when i listen to it i'm gonna be like oh why why did i talk about marshmallow peeps every five seconds i wouldn't be surprised if there wasn't a peep about it in the final cut so ernest let's review what we've learned through this episode i've actually picked up on a lot in this interview, and, but also how does it move?

How does it interact with other cells? And you really cannot see that without supercomputing. It's true. I mean, while it's not cool that the pandemic is in our midst right now, it's better now than it would have been before this technology existed. And imagine what it was like in 1918 when the last pandemic came. Yes. We've learned about how these sugars are disguising the virus to look like any of our other human cells. And it just goes to show that, I mean, every virus is different out there.

And I'm sure there'll be another one that's going to be clever and sneaky, like coronavirus coming down the pipe. So I'm grateful for the scientific research that we have and the supercomputing that can take us to new heights. So if you want to learn even more about these studies, you can check out the episode notes on bigcompute.org where we'll drop some links and some pictures and such for you. You can also go to amarolab.ucsd.edu for more technical information. And you can follow Romy on Twitter via at Romy Amaro. And that's R-O-M-M-I-E, two M's, Romy Amaro. And that's R-O-M-M-I-E, two M's, Romi Amaro. And we invite you to follow Big Compute on social media on LinkedIn, Twitter, Facebook, and YouTube. Yes, all of our social media accounts

are somewhat new. And we're especially brand spanking new on YouTube, like literally a few days ago. So please go show us some love. I think we have nine subscribers and they're probably all my mom. And we wanted to give special thanks to Romia Morrow and her awesome team at UC San Diego, as well as Dan and TACC, the NAMD software creators, the COVID-19 HPC Consortium, Tech Against COVID, and the teachers and people who have inspired these researchers to do what they're doing. Yes. Oh, in fact, there was a professor Romi mentioned by name who inspired her to study chemistry when she was an undergrad. And his name was Steve Zumdahl. So like these scientists, researchers and engineers, we're grateful for the teachers

out there who are making a difference. And to all of you out there who are conducting research or on the front lines of this virus, a special thanks and shout out to each of you from us at Big Compute. That's going to do it for this episode of the Big Compute podcast. I think this is where we insert a plea to leave us a five-star review, but I feel stupid asking for it. I'll do it. Thank you. Leave us a five-star review on Apple Podcasts. Yay!

Or wherever you listen to your podcasts. See you next time. Bye! Thank you.