AI-designed antivenoms could help treat lethal snakebites
33:52 Share

Nature.

Welcome back to The Nature Podcast. This week, a computational method for making life-saving anti-venoms. And news that Earth's temperature has breached a significant climate threshold. I'm Lizzie Gibney. And I'm Benjamin Thompson. There's a neglected tropical disease that affects millions of folk around the world and is estimated to kill over 100,000 people each year and cause long-term health impacts for many more.

The WHO listed it as a highest priority neglected tropical disease back in 2017. Now, I'm sure you're trying to figure out what it might be, and the answer may surprise you. It's snake bites, specifically the bites of venomous snakes of various species. Snake envenoming is a significant health burden, and that burden is highest in the poorest countries of the world, especially in remote rural settings.

One of the reasons for this is that treatment options are limited. Currently, the only approved way to treat snake venom is with the rapid use of antibody-based antivenom. And these antivenoms can be difficult and expensive to make and tough to get to the places where they're needed.

To potentially overcome some of these issues, a team has come up with a method of designing new antivenoms from scratch with the help of computational machine learning, and they've a paper about it in Nature This Week. The team put their method through its paces by designing potential antidotes against what are known as three-finger toxins, produced by elopid snakes such as cobras.

To find out more, I spoke with lead author of the paper, Susana Vázquez-Torres, from the University of Washington in the US, who laid out what these three-finger toxins can do to the human body.

So these three finger toxins affect the nervous system, which leads to paralysis. And also they have cytotoxins, which could lead to necrosis. So you could imagine that the effects they have in the body are very, very bad. But venoms are complex mixtures of different types of proteins and peptides and enzymes. It seems like that the current treatments for bites by Ellipid

snakes i mean there's only one option really and that's antibody therapy the current treatments are actually based on purified serum for immunized animals so that means that you could take whole venom of medical relevant snake and then you inject it to a horse for example they

immune system of the horse raises antibodies against different components, and then the blood is extracted, and these antibodies are purified and injected to patients. I mean, as you can imagine, it's not very good to rely on horses, especially because we have better ways to make better treatments right now. And also, as this problem affects developing regions,

You can imagine this could be expensive. Also, they need to be like on a fridge, which is not always possible in rural communities. So even though they are effective, they face a lot of challenges. They could also be dangerous because besides injecting neutralizing antibodies, you are injecting horse serum proteins. Our body, it's not a horse. So basically our body will fight against these foreign proteins. And antibody-based drugs.

antivenom therapy has been the go-to standard then for like over 100 years. You and your colleagues have taken a different approach. You've used computational deep learning to design new small protein antivenoms. Why did you think this is a good problem to throw at this computational technique? Yeah, a lot of people have been trying to improve the current treatments by refining just the neutralizing components on the horse serum or

developing nanobodies, which are like a smaller version of antibodies. But all of these options rely on the immune system on an animal, right? But with recent advances on deep learning, you don't rely on the immune system of any animal because we have very powerful tools to immediately develop proteins that can bind to these toxins and test them in the lab so we could accelerate the discovery process compared to the other methods.

And it also offers some advantages as the novel design proteins are known to be very cheap to produce, very heat stable. And also because they're very tiny compared to antibodies, they could potentially penetrate tissues in a faster way. And you were looking for molecules that worked against a

a couple of different types of three-fingered toxins. We mainly focus on alpha neurotoxins that basically can block the muscular system or the nervous system. And we are also targeting cytotoxins with disrupt cellular membranes that could lead to necrosis or bleeding. So we decided to limit ourselves to very important classes of

toxins and design proteins to see if we could neutralize them and it could have some effect on mice, for example. So this is the first step towards like a real fully de novo design antidote for a snake bite. So you were working in silico, then working in the computer space to try and design these small proteins. And what was the kind of workflow you did? Did you start with just a blank slate and work from there? So the first step is just to get a 3D structural information about the toxin you want to neutralize.

So basically, we would like to have proteins that bind very specifically and very tightly. In this case, we try to...

like match their shapes with another protein. And in order to generate these shape match proteins, we use an algorithm which basically enables us to generate a lot of like complementary protein backbones. This target, just like the skeleton of a protein. So after that, you need to use a second program that assigns an amino acid sequence to those skeletons.

And then to evaluate the quality of your designs, you can use algorithms like AlphaFold. Like it tells you if it's good or bad. Because in this process, you generate hundreds of thousands of designs. So imagine it's very difficult to by eye filter all of them. But once you filter this set of designs, you get like smaller set that you can visually inspect. You need to choose things that the computer think they look good, but you also need a chemist to make sure they look good. All right. So you have your venom protein sample.

structure and you can see in kind of 3D space there's a gap there that if we jam that gap we might stop it working you can get a skeleton to fit into that gap and then you need to put the amino acids on the skeleton and then see would that make a real protein is that about right? That's exactly correct So your computer then spits out its candidates then of most likely and then you yourself make those in the lab test that they actually do bind to

to the toxin proteins. But I suppose, you know, seeing that it does that is one thing, but actually making sure that it works, that it's efficacious is quite another. And I know that in your work, you had to turn to animal models to see what happened. And you tested a couple of molecules that were designed against

against two separate alpha neurotoxins. What did you see? I mean, that was very shocking because I think most of the novel protein designers are kind of used to see like things work on a petri dish. So for us, it was really like...

A very exciting thing to see that we basically injected mice with lethal doses of purified as neck toxins at a concentration that we know that for sure all of the animals will die. And then we treat them with our design antidotes and measure how many of those mice survive after 24 hours. Most of the cases were positive, but in some of the cases it was 100% protection, which was very shocking. And time, of course, is of the essence if someone is bitten.

by a venomous snake and you did different time courses to check the efficacy of this what did you see in that aspect i mean yeah of course we tested like conditions where we kind of mix our antidote with the venom which doesn't represent reality because people get bitten and then they get treatment so in this like artificial condition a hundred percent of the animals survive and

in all of the cases. But when we tried to challenge the animals first to a lethal dose of toxin, and then we waited 15 minutes after that, also 100% of the animals survived, which was also like very shocking. And we tried to increase the time to 30 minutes. And for one of our toxin targets, we were able to achieve 100% protection. And the other toxin was also neutralized, but a little less, but a lot of the animals were protected.

And this work was, of course, in mice and a small mouse study. What happens to these molecules now? I mean, I think the most obvious next step would be to start performing some pharmacokinetic studies to understand how these proteins actually enter and react in the body. So we could design a product that can go into humans.

I think we need to generate more neutralizing proteins for other snake toxins so we can make a cocktail of the same proteins for whole venom. Because in this study, we only neutralize purified toxins, which doesn't represent reality because a snake has a mixture of different things.

And when the moment comes, try to jump into clinical trials. And there are other issues that have to be overcome too, it must be said. One of the compounds you tested didn't show the same efficacy as the two you've discussed there. What does this work not do? As I say, I don't think this work solves the whole problem, but I think it brings...

a lot of novelty to the field. I think it's just the very, very first step, but I think it's exciting because we're bringing attention to the scientific community, but I also hope to pharmaceutical companies to bring attention that a snake bite is a huge problem and we have better ways to do it. We just need to come together. And is it the molecules themselves or is it the method you used previously?

to create them that you think is most important in this work? I think the methods, I think the neutralizing proteins we design are efficacious and serve as a proof of concept. But I think deep learning methods for protein design and structure prediction methods will really accelerate the discovery of more molecules like these ones. And finally then, how do you hope that this in future could be used?

I mean, I think this could just bring the attention of more scientific people out there and also maybe government and communities. So we could actually work together and actually decide how we can do this to make a real life treatment. It's not trivial to decide which species of snake we need to target, but I'm sure if we come together to define what are the things

that need to be solved, we could make it happen very quickly. So we could maybe in a couple of years come with a product that actually could save some snake bite victims in, for example, South Saharan Africa, which is like a region that faces a lot of snake bites. Susana Vázquez-Torres there. To read her paper, look out for a link in the show notes. Coming up, news that in 2024, average global temperatures exceeded 1.5 degrees Celsius for the first time, alarming scientists around the world.

Right now, though, it's time for the research highlights with Dan Fox. For male wasp spiders, finding a mate is a case of following your nose, or rather your legs. That's because these arachnids have been found to use hair on their limbs to sniff out potential partners.

A sense of taste has been documented in many spiders, but how most species smell airborne chemicals remains a mystery, despite evidence that a number of them use such chemical signals to hunt, evade predators and find mates.

Now researchers have used electron microscopes to analyze the appendages of wasp spiders. They found pores in sensory hairs on the legs of adult males, and experiments showed that neurons in these hairs reacted to a pheromone released by female spiders to attract mates. The team identified similar structures on the legs of male spiders from 12 different species, but also found that some other species didn't have them.

The authors say this suggests these organs may have evolved independently more than once in spiders. Sniff out that research in the Proceedings of the National Academy of Sciences of the United States of America. As mice settle down for a good night's sleep, their brains are doing more than just dreaming of cheese. A new study shows the organ gets busy cleaning itself, flushing out waste products that have built up during the day.

Researchers monitored the brain activity of mice when they were awake and asleep. They found that during deep sleep, a particular brain region released noradrenaline, a molecule which causes blood vessels to constrict about once every 50 seconds.

This made blood vessels in the brain tighten and relax in rhythmic patterns, powering the brain's natural cleansing system by pumping cerebrospinal fluid into brain tissues which carries off waste. The researchers also saw that when given the sleep aid Zolpidem, this process vanished, reducing the flow of CSF by 30%, possibly decreasing the clearance of waste proteins.

Don't sleep on that research. Read it in full in Cell. News broke at the end of last week that in 2024, Earth's temperatures breached a significant climate threshold, exceeding 1.5 degrees Celsius above pre-industrial levels.

Jeff Tollefson has been covering the story for Nature, and he joins me now. Jeff, hi. Oh, Ben, good to be here. Good to have you with us. And it does seem like this event does have the potential to be one of those things that's sort of marked down in the timeline when we look back at the story of climate change. Before we get into the context...

Maybe we should define what we mean by this 1.5 degree threshold having been broken. How was this calculated? And what are we actually talking about when we say 1.5 degrees? Yeah, that's a very good question. And we need to be specific about it here. There are different ways that you can measure things. And there are different ways that climate scientists measure things. In this case, what we're talking about is the global average surface temperature over the course of one year. So 2024 in particular.

Climate scientists created this metric a long time ago as a way to try and kind of basically, literally kind of take the Earth's temperature and monitor the Earth's temperature in the same way that you might a patient in the hospital. And so there are several kind of climate services that measure this temperature in different ways across the planet, and they come up with their own calculations. In this case, the World Meteorological Organization, you

ran the numbers, averaged those different climate records together and came up with a formal tally for 2024 of 1.55 degrees above pre-industrial levels, which is defined as the average from 1850 to 1900. But that's on a one-year basis, right? That doesn't necessarily mean anything about the future. And this limit is not actually defined in

in the 2015 Paris Agreement. Yeah, lots to unpack there. But I will say that this 2024 temperature follows on the back of 2023, which some researchers were saying sort of nudged the 1.5 threshold. But here we are, 1.55. And it's these fractions of fractions of degrees that are really important. I mean, these seem like small numbers, but you have to keep in context what they represent. They are an indicator for much broader changes of climate across the planet.

And they're also an average. So these numbers kind of average out all of the crazy impacts and all of the extremes that we all are familiar with these days.

So 2024 comes on the backs of 2023, which was itself a scorching year. It broke all records. And in fact, it stepped outside of the envelope of what climate scientists consider their average projections going forward. So it was a statistical anomaly. It was unusual. So unusual that climate scientists are trying to understand what happened and why they feel like they need to provide an explanation.

So 2023 and 2024, we now have two years in a row that are just extraordinarily warm. And scientists don't have a perfect answer for what is causing this warmth. And the question becomes, you know, is it just random climate variation? Climate varies year to year in the same way that temperature varies day to day. Or does it signify that actually global warming is accelerating?

And that, you know, if that's true, then we should expect even more warming than we've been warned about in the future, in the decades to come. Yes, I imagine this. Folks will say, well, it was an El Nino year. This is exacerbated or caused a blip. It is true that 2023, we were in the midst of kind of another record-breaking El Nino, this ocean event where the

equatorial Pacific Ocean, basically it warms up. You get this mass of warm waters on the surface. And when that happens, the global weather patterns kind of go crazy. And in particular, the average global temperature tends to be warmer. So this is one explanation for what happened in 2023. And the expectation was that temperatures would drop once that El Nino ended, and they did not. So the El Nino alone does not seem to be enough to explain what happened in 2023.

And we have no explanation for why 2024 was as warm as it was when El Nino ended. And what are climate scientists saying in the face of this announcement then last week that this 1.5 has been broken? At the very top line, it is alarming. This is one of those landmark moments where you have to just kind of hit pause and ponder the implications.

What it tells us is that the temperatures are rising and they're rising faster than we thought. And that is cause for alarm all by itself. 1.5 degrees, of course, it's not this magical threshold. Above that, everything's terrible. And below that, everything is safe. We know for a fact that today there are all sorts of climate disasters happening. And officially, the world is at 1.3 degrees on a long term average, not the annual basis, but on a decadal basis.

So if the world is dangerous at 1.3 degrees, that means it's going to be even more dangerous in five or 10 years when we breach 1.5 degrees on a decadal basis. And averages then are important, as you said a few times. How does this 2024 result fit into the Paris Climate Accord from 2015 to keep global warming to 1.5 degrees above pre-industrial levels? So 2024 was one year.

There are different ways that you can measure this. As I said before, the typical way these days is that scientists use a decadal average. And by the decadal average, we're at 1.3 degrees of warming today.

And if you just kind of assume that temperatures are going to bounce around a little bit and maybe continue to rise a little bit based on current trends within several years or five or 10 years, almost certainly we're going to breach 1.5 degrees on a decadal basis for the first time.

There's no formal definition for what the 1.5 degree limit, how to define that in the Paris Agreement itself. Countries, when they signed that agreement, they basically left it vague, which means it's kind of up to scientists for how to define it. And so a decade-alverage seems to be what people generally accept.

So we are likely to breach that even on a decadal basis. But keep in mind that, you know, for some time it's been clear that we're likely to go over that 1.5 degree limit temporarily at a minimum. But a lot of the models that are used by climate scientists, the 1.5 degree trajectory models, the ones that we use to kind of analyze what a world might look like if you were able to meet the Paris goals,

A lot of those models assume that humanity goes above 1.5 degrees, that greenhouse gas emissions push global temperatures above 1.5 degrees now. And then later in the century, we pull CO2 out of the atmosphere and the temperature cools back down by the end of the century.

So in theory, you can still meet the Paris goals, even if we have a temporary breach of 1.5 right now. And do you think there's a possibility that this significant 2024 event that researchers are alarmed about actually could be useful in that it sharpens

government and policymaker sort of minds, as it were. Of course, we've got COP coming up in Brazil this year. Could this be potentially a turning point in how this is dealt with, do you think? There's no way to predict that. What I can say is that there are a lot of scientists and environmentalists who will go to the big UN climate conference and use this as yet another warning for governments as they urge them to take additional action. And

and to take more aggressive action and to speed up the efforts that they have promised to undertake. Keep in mind that it's not just what happens at the big climate conferences, at the COPs, that matters. Ultimately, you know, government leaders have to leave these COPs and come home and make decisions that will push emissions down, whether that's regulating emissions or putting other policies in place that will encourage clean energy in place of fossil energy.

So there's a lot to be done. Will this add political pressure on leaders? Certainly. How they will respond, I have no idea. Well, here we are at the start of 2025, and I'm sure researchers are going to be very, very keenly looking to see what happens in terms of working out whether these last couple of years were abnormal or whether, as you say, things are speeding up.

But the start of this year, of course, seen severe events. I'm thinking about, of course, the fires in Los Angeles that many researchers are saying that events like these are going to become more frequent and more intense as time goes on. Absolutely. The wildfires in L.A. are...

you know, just terrifying. And it's yet another reminder that we don't necessarily live in a safe world just because we're below 1.5 degrees C. Climate change is already throwing any number of things at us from increasingly severe and frequent wildfires to floods to droughts. All of this is happening today while we are in a supposedly kind of safe zone.

It's good to keep in mind that as the temperatures increase, these types of things are only going to get worse. And I think that's the scariest thing for scientists. On the bright side, perhaps one scientist I spoke to said that we need to listen to nature, that nature is our teacher. So these signals, if we read them correctly,

They will help us understand the impacts that are here today and the impacts that are coming. So maybe if we listen, we will start to take our climate commitments more seriously and we will start to turn the corner a bit faster on climate. Nature's Geoff Tollefson there. To read his story, look out for a link in the show notes. Finally on the show, it's time for the briefing chat, where we discuss a couple of articles that have been highlighted in the Nature briefing.

Ben, what have you been reading this time? Well, I've got a story that I read about in Nature that we've been following for quite a while here at Nature and on the Nature podcast as well. And it's about NASA's plans to bring Mars samples back to Earth. And there is an update. OK, the last I remember, it was looking like it's going to be very, very, very expensive. Oh, eye water and expensive was what the last estimate was, like £11.

billion dollars but let's set the scene so of course nasa's perseverance rover has been for a few years now collecting samples of dust and rock from the martian surface it landed in the jezero crater and it's been sort of trundling around there hasn't it sort of collecting stuff a lot of those samples are in its belly it's got a few empty sample containers to go but it's been picking them up i think it's just outside the crater now but the question has always been now what how do we get them back to earth and as you say numbers were thrown about that were just

Well, out of this world, he says, some pun intended. And a 2023 review estimated, yes, this amazing cost, which was obviously no good for a cash-strapped agency. And so last year, I think it was early last year, they said, this is no good. And another report said it's going to take to the 2040s, which is too long. So NASA said, all right, let's open up proposals to do it quicker, cheaper, faster, whatever it is, these sorts of things. But it wasn't just fast.

to NASA sensors. It was to private companies as well. And then we got an update this year, so the 7th of January. And they said that something would happen. And as I say, I've got an update to give you. Okay, so you're leaving me on tent hooks. What is the update? The update is they're going to think about it and get back to us next year, Lizzie, I must confess. So yeah, it's one of those things that you get, I guess, one go to do this right. And so

The good news is that they're actually looking at two options for the mission, right? So they've actually narrowed this down, one using this well-tested NASA technology and one relying on systems being developed by private companies. Now, the earliest that either of these could be done, I understand, is going to be 2031 and samples back by 2035. So earlier than the previous plans, I suppose. And both of these come in at the sort of six to seven billionth.

billion dollar range so obviously significantly less but obviously still significant yeah so earlier and cheaper so that's that's kind of good news but private companies seem to be the answer a lot in the space game at the moment right yeah and there are a lot of them but i think what

what's interesting here is both the NASA plan and the private plans are kind of similar. A spacecraft launches from Earth, delivers a lander onto the Martian surface. The Perseverance rover comes up, meets that lander, drop the samples off, lander then blasts back into orbit. And then another spacecraft from the European Space Agency picks that up in this kind of giant relay race, takes the samples back to Earth. Seems really, really straightforward, right? And the

this is where the decision needs to be made. So the NASA version relies on what's called a sky crane, which is what was used to drop the Perseverance, well, to gently put the Perseverance rover onto the surface of Mars. It is essentially a crane that hovers in the air. It's been done before.

The private company plan is really not known at the moment. It seems that it will be a system of large rockets and landing mechanisms and details are sparse. But whichever way they go, this mission is likely to be cheaper than previous proposals because the rocket that takes the samples from the surface to orbit is smaller and cheaper to transport. And obviously, you know, smaller means less weight, but all the rest of it.

And so I know that Perseverance has some incredible samples, you know, some really tantalizing findings that really could do with being confirmed in lab back on Earth. So it's a very, very important mission, but very expensive. What happens next? When do we know if we know if this is even going to happen at all and how? Yeah, that's an important question. And you're right. Scientists are desperate to get their hands on this stuff. The rovers that have roved around on Mars have taught scientists so much about this planet and its potential history and all the rest of it.

But they can only do so much. And so getting hands on and looking at the soil is really, really important. And it has to be said that, you know, they aren't the only people trying to do so. China are trying to do the same thing. Right. And reports saying that maybe as early as 2031, they could bring back samples. But obviously, that's to be discussed. But in terms of what happens to the NASA attempt, yeah, lots of questions there. Of course, funding is the big issue. We've got a new presidential administration coming in who have proposed and so have the incoming Republican led Congress.

congress to slash budgets across the federal government and six to seven billion dollars is a lot of dollars so maybe let's chat again next time when the things come out and we'll continue this conversation but in the meantime let's move on lizzie what have you got for me this week so from you know these very big things in space to very teeny tiny things in the biology world

So this is a story in Nature about how the world of studying bacteria is incredibly skewed towards maybe 10, 20 very, very, very popular microbe species. And of course, that could be a problem for many reasons. So a microbiologist called Paul Jensen at the University of Michigan in Ann Arbor said,

tried to look at just how skewed the studies are. So he looked at a database of 43,000 odd species and counted up the number of those that had appeared in PubMed publications, so the repository. And he found that the top 10 accounts for 50% of all papers, right? So just 10 out of this 43,000.

It's 50% of all papers. And that almost three quarters of species were not mentioned at all. Not in any title or abstract of a single paper.

So what do you think are in the top 10, first of all? Give me some of your top hits. Okay, so as a former microbiologist, I want to say E. coli is going to be number one by a mile. Ding, ding, ding. E. coli is number one. It's got 21% of the total, so that's some 300,000-odd papers. Right, okay, so that's number one. Yeah, understandable. So I guess they're going to be skewed towards human pathogens, so let's think what a significant one. I guess tuberculosis is one of the biggest ones

bacterial pathogens so mycobacterium tuberculosis that's right that's number four okay oh right okay streptomyces coelacolor which is the bacteria that i used to work on no staff yeah some pseudomonas um other ones people might have heard of helicobacter pylori which is the cause of stomach ulcers people might know in that top 20 you've also got the bacteria behind chlamydia gonorrhea you've got salmonella listeria all your favorites absolutely so the

there are a lot there that are obviously very, very important, but there are a lot which are not there. Now, one question I had, which actually isn't the story and I thought maybe I would ask you. Right.

is if we've got all these species that appear in no title or abstract in PubMed, how are they even species? Like, they must have been looked at in some capacity. Yeah, that's an interesting one. So for a species to be named, it has to be described, as I understand. So you've got a whole podcast about that. Well, I did have a whole podcast about that. What's in a name? Yeah, so that's interesting. So I wonder if they've been given sort of provisional names and then data on their genome has been uploaded to the website. But I mean, I wonder...

Why did he do this in the first instance? Well, he had a particular microbe that he was studying and he thought, why don't I do what everyone's doing right now, apply AI to some data on this microbe and see what happens and see what I might be able to get out of that from pooling all of this data together.

And he found that that was basically impossible because there were only a few dozen papers on this Streptococcus Sobrinus, which is an organism that causes tooth decay. And he actually had read all of them already. So he realized when he was doing this that actually a lot of bacteria out there are very, very understudied.

So this is a problem if we are trying to do exactly what he was trying to do, apply AI to data. If there is no data, you can have no model, you know, is the absolutely fundamental ingredient. And of course, it's just a problem in general. A lot of these that we're missing are in the human microbiome. They are in the sea, they're in the soil, they interact with each other. They interact with some of the ones which we know are very prevalent and important in humans.

And we just know very little about them. So, you know, there are some efforts underway to change that. Paul Jensen is actually part of this Align to Innovate project, which is trying to characterize the growth of a thousand microbes in a thousand different individual conditions. So that will obviously bolster the amount of data available.

that's out there but for now yeah E.coli reigns supreme and of course E.coli is the model organism right so there's a reason why it tops the chart. And also it's super easy to work with as well anyone who's worked with E.coli you can set up your overnight culture and then come to work the next day and have enough to work with it grows at a sensible temperature you know all this kind of stuff right it's easy to fiddle about with genetically but yeah it's an interesting one right because we could swab this desk in front of us here and probably find a new species of bacteria and

it would go in a database and then nothing else would be done with it. So yeah, it's a fascinating one, really. And I guess I wonder how he feels as well, that like he's read every single piece of work associated with the thing that he studies, but has done something about it. Yeah, that's it. I mean, it must be a very strange feeling. Maybe a nice feeling. I don't know. You know everything about the field of study. But he's also, he's optimistic that

that scientists are going to be able to learn to grow these other kind of important but understudied microbes. Well, that is a terrific story and such an interesting one that I'm covering space, you're covering microbiology, although a bit with an AI bent. Up is down, down is up. I mean, I'm not sure where the podcast is going now in 2025. Absolutely. We just turned it on its head. It's fine by me. Right. Well, listeners, for more on those stories and where you can sign up to the Nature Briefing to get more like them delivered directly to your inbox, check out the show notes for some links. That's all for this week.

As always, you can keep in touch with us on X, we're at Nature Podcast, and soon look out for us on Blue Sky. Or of course, you can send an email to podcast at nature.com. I'm Lizzie Gibney. And I'm Benjamin Thompson. Thanks for listening.