Rapid sepsis test identifies bacteria that spark life-threatening infection

Primary Topic

This episode discusses a groundbreaking rapid diagnostic tool for identifying bacterial species in sepsis cases, potentially transforming patient treatment.

Episode Summary

In this compelling episode of the Nature Podcast, the focus is on a revolutionary diagnostic tool designed to drastically reduce the time required to identify bacterial species responsible for sepsis and determine their antibiotic sensitivity. Sepsis, a critical condition where the body's response to infection causes tissue damage and organ failure, is notoriously difficult to diagnose and treat effectively. The new testing method, developed by a team led by Seung hoon Kwon at Seoul National University, can identify bacterial infections in blood and ascertain antibiotic susceptibility in under a day, bypassing the traditional multi-day culture process. This advancement could significantly improve treatment outcomes by allowing healthcare providers to administer the most effective antibiotics sooner.

Main Takeaways

  1. Rapid Diagnosis: The new test can identify bacterial infections in less than a day, a significant improvement over current methods.
  2. Improved Treatment: By quickly determining the most effective antibiotics, this test could drastically reduce mortality rates associated with delayed sepsis treatment.
  3. Reduction of Broad-Spectrum Antibiotic Use: The test allows for a more targeted approach, reducing the reliance on broad-spectrum antibiotics and helping to prevent antibiotic resistance.
  4. Technological Innovation: The test uses magnetic nanoparticles to isolate bacteria from blood samples, showcasing an innovative approach to medical diagnostics.
  5. Clinical Integration Challenges: Despite its potential, integrating this new technology into existing medical workflows remains a significant challenge.

Episode Chapters

1. Introduction to Sepsis

Overview: The episode begins by discussing the severity and complexity of diagnosing sepsis. Ron Daniels: "Sepsis is a really hard condition to recognize and diagnose."

2. The New Diagnostic Tool

Overview: Details the development and functionality of the new rapid test for sepsis. Seung hoon Kwon: "We have a little magnetic nanoparticle... that can capture the bacteria."

3. Clinical Implications

Overview: Explores the potential impacts of the rapid test on clinical practices and patient outcomes. Ron Daniels: "It's much more rapid, it can eliminate steps that many other technologies necessitate."

Actionable Advice

  1. Stay Informed: Keep up-to-date with the latest advancements in sepsis treatment and diagnostics.
  2. Advocate for Modernization: Encourage healthcare facilities to adopt new technologies that improve diagnostic speed and accuracy.
  3. Educate on Sepsis Awareness: Increase awareness about the signs and symptoms of sepsis to promote early intervention.
  4. Support Research: Contribute to or support research initiatives aimed at improving sepsis diagnostics and treatment.
  5. Engage with Healthcare Professionals: Discuss new diagnostic tools and their potential benefits with healthcare providers.

About This Episode

A newly-developed method that can rapidly identify the type of bacteria causing a blood-infection, and the correct antibiotics to treat it, could save clinicians time, and patient lives. Blood infections are serious, and can lead to the life-threatening condition sepsis, but conventional diagnostic methods can take days to identify the causes. This new method does away with some of the time-consuming steps, and the researchers behind it say that if it can be fully automated, it could provide results in less than a day.

People

Seung hoon Kwon, Ron Daniels

Companies

Seoul National University Hospital

Books

Leave blank if none.

Guest Name(s):

Leave blank if no guest.

Content Warnings:

None

Transcript

Nature Podcast
The Nature podcast is supported by Nature Plus, a flexible monthly subscription that grants immediate online access to the science journal Nature and over 50 other journals from the Nature portfolio.

More information at go dot nature.com plus.

Ryan Reynolds
Ryan Reynolds here for, I guess, my hundredth mint commercial. No, no, no, don't, no, don't. No. I mean, honestly, when I started this, I thought I'd only have to do like four of these. I mean, it's unlimited to premium wireless for $15 a month. How are there still people paying two or three times that much? I'm sorry, I shouldn't be victim blaming here, give it a try@mintmobile.com.

switch whenever you're ready.

Unknown
$45 upfront payment equivalent to $15 per month. New customers on first three month plan.

Unknown
Only taxes and fees extra speeds lower.

Unknown
Above 40gb cd tails.

Ron Daniels
In an experiment.

Unknown
We don't know yet.

Unknown
Why is blight so far like it sounds so simple? They had no idea, but now the data's I find this not only refreshing, but, but at some level astounding.

Ryan Reynolds
Nature, nature welcome back to the Nature podcast. This week, a faster way to help.

Unknown
Doctors treat sepsis, and how giving AI AI generated data can make things go strange. I'm Emily Bates.

Ryan Reynolds
And I'm Nick Perchelle.

First up this week, reporter Benjamin Thompson has been learning about a diagnostic tool that could speed up the treatment of patients with a life threatening blood infection.

Benjamin Thompson
Sepsis is a serious condition where the body's own immune system essentially goes into hyperdrive, damaging organs, potentially leading to death.

It's incredibly common, affecting tens of millions of people each year, and is a leading cause of mortality around the world, but can be really difficult for clinicians to pick up.

Ron Daniels
Sepsis is a really hard condition to recognize and diagnose.

Benjamin Thompson
This is Ron Daniels, an intensive care doctor and joint CEO of the charity the UK Sepsis Trust.

Ron Daniels
It can affect people of any age. It can arise as a complication of any infection anywhere in the body, and people present in a myriad of different ways. Yes, there will be some people who present in obvious threatened or actual multi organ failure, but for every one person like that, there will be a dozen people who present with a much more insidious onset.

Benjamin Thompson
And while some people may present with less serious symptoms, it often doesn't take much time for them to get into trouble.

Ron Daniels
And that's what we have to look out. For now, there's very variable estimates around the impact of delays in recognizing and treating sepsis, and the estimates for each hour's delay range from about a 1% increase in mortality to, at the greatest end, an 8% increase in mortality, and the true impact is likely to be somewhere in the middle. So this is a medical emergency. Yes, it's insidious, but that doesn't mean we don't have to act quickly.

Benjamin Thompson
And part of acting quickly means getting the right treatment.

While sepsis has many causes, one of the leading ones is bacteremia, a bacterial infection of the blood.

Typically, treatment involves the rapid use of broad spectrum antibiotics. These are vital generalist drugs that don't target a specific bacterial species, and they're used to try and get on top of an infection while doctors try and figure out a what it is. But extended use of these drugs can increase the chances of resistance against them developing. They can have side effects, and they might just not be the most effective antibiotic available.

Ideally, clinicians want to limit the use of these broad spectrum antibiotics and quickly move a patient onto narrow spectrum antibiotics. But working out which ones are effective can in many cases take several days, increasing the risk of adverse side effects and potentially death. But this week, a team has designed a new way that could, if fully realized, drastically reduce the time it takes to identify the species of bacteria involved in an infection and the antibiotics they are sensitive to, like 30 to 50.

Seung hoon Kwon
Hours less than the conventional.

Benjamin Thompson
This is Song hoon Kwon from Seoul National University in Korea, one of the research team behind the new work.

Songhoon has been working for years to develop ways for clinicians to quickly diagnose whether someone has a bacterial blood infection and what's causing it.

The conventional way to do this takes so long, because while a patient's blood sample can contain millions upon millions of red and white blood cells, at the early stage of an infection, it might have fewer than 100 bacteria, not enough to run tests on. So the first step is to increase this number.

Seung hoon Kwon
We put this blood into media where bacteria can help it to grow, and then we raise the temperature for at least more than ten to 24 hours.

Benjamin Thompson
This growing step, known as blood culturing, is important because if nothing grows, the patient's condition is not bacterially based.

If it does, though, there are more steps to come. A second pure culture step is required to remove all the blood cells before the bacteria can be identified, and an antibiotic susceptibility test, or an ast, can be done to see which drugs work. And together, all this takes time.

Seung hoon Kwon
So it takes like three, four days from the blood draw to get the optimal antibiotic result.

Benjamin Thompson
So to help get survival rates up, researchers have been working to get this analysis time down some rapid antibiotic susceptibility tests, known as rapid asts, do exist, and Song hoon has been involved in commercializing some of them. Rapid asts work in different ways, but can save about a day by doing away with the pure culture step. However, they still require the initial blood culture stage.

In this work, Songhoon and his colleagues have developed what they call an ultra rapid ast that can do away with this blood culture step as well. And they estimate it could produce results in less than a day.

So how does it work? Well, remember how those initial blood samples taken from patients could have an incredibly low number of bacteria? Well, this method scoops them all up.

Seung hoon Kwon
The principle is simple. We have a little magnetic nanoparticle. In the surface of the magnetic nanoparticle, we coat the peptide that can capture the bacteria.

Benjamin Thompson
These peptides stick to a wide range of pathogenic bacteria. Once attached, a magnet can be used to separate them out from all the blood cells, a process that takes around an hour, doing away with the need for a lengthy blood culture.

The purified bacteria are then used for a genome based species identification technique or grown in culture for a few hours.

This increases their numbers and allows them to be placed in essentially tiny petri dishes on a custom made growth plate designed to rapidly test test their sensitivity to different drugs.

Seung hoon Kwon
This bacteria goes into many different chamber with many different antibiotics in different concentrations, and then we image them. So we look at the bacterial growth pattern for a few hours to know this ast result faster.

Benjamin Thompson
In the lab, the teams method showed good results at assessing antibiotic sensitivity compared to conventional testing.

The next step was to measure the accuracy of their approach in a clinical setting. So they enrolled 190 patients in Seoul National University Hospital who were suspected of having a blood infection.

Seung hoon Kwon
So basically we take out the patient blood, the one bottle goes to the conventional workflow and the other bottle goes to our workflow and we compare it.

Benjamin Thompson
About 10% of those involved tested positive for an infection, so the team looked to see how their methods stacked up in terms of species identification.

They also tested their antibiotic sensitivity assay on six patient derived bacterial strains.

Seung hoon Kwon
So we got 100% accuracy. On IED researcher for the antibiotic result, we got more than 90% accuracy.

Benjamin Thompson
Seung hoon estimates that if their method is perfected, antibiotic sensitivity tests could be turned around in less than a day from initial blood taking.

Ron Daniels, the intensive care doctor you heard from at the start, was impressed.

Ron Daniels
I think the most important thing about this particular technology is it's much more rapid, it can eliminate steps that many other technologies necessitate. So it improves processing time. It might, in time, pave the way to bringing the technology closer to the patient. It's much more sensitive in terms of genetic identification of the bug. It can be a thousand times more sensitive than other methods that don't rely on a pre culture of blood. So, yes, it's one of a number of technologies in this space that are emerging into the marketplace, but it is one of the more exciting ones.

Benjamin Thompson
Ron says that this rapid speed could mean that his patients at most need two doses of broad spectrum antibiotics before being placed on a targeted narrow spectrum, one which has numerous benefits to patient outcomes.

He also says that preventing the overuse of broad spectrum antibiotics should help protect these drugs from the threat of resistance. However, Ron highlights something that all emerging antibiotic sensitivity tests face, fitting them into workflows.

Ron Daniels
The technology around diagnostics is evolving at an infinitely greater rate than our ability to integrate it into clinical systems, and therein lies the knowledge transfer gap. We've got to get a lot better at building this technology into clinical systems.

Benjamin Thompson
For Ron, having a rapid or ultra rapid test is one thing, but if it takes a day for a sample to get to the testing facilities, and the results take a further day to get back into his hands, any benefits of speed are lost. So he says that figuring out the best way to get systems working together as efficiently as possible is critical for patient care. Seunghun is acutely aware of this too, and knows that working with clinicians and hospitals to streamline processes is key for technologies like this to be a success. And these are hurdles that will need to be overcome. And there are others too. At the moment, this method has many manual steps. Seung hoon wants to work to combine and automate it all into a single device, and if it passes the regulatory tests, it could be used in the not too distant future.

Seung hoon Kwon
I think that this technology can be in one box within three years, and then one more year will be needed for regulatory.

So if we have enough funding, then within four years it can be in.

Benjamin Thompson
The clinic, and if it gets there, it could help improve the lives of patients with sepsis, their loved ones and their doctors.

Ron Daniels
So for the individual patient, having this rapid diagnosis means that they and their families, they might be unconscious. So their families, perhaps can have confidence that we now have the information. We know what the bug is, we know what it's sensitive to. We're now delivering exactly the right therapy that's going to treat them, whilst minimizing the unintended consequence of treating them. But I think there's another aspect. There's always health professional anxiety and skepticism around any missive that says we've got to treat something rapidly using a potentially dangerous agent. If the health profession has confidence that we have technologies like this that guide us to deliver the right therapy properly, it will make improvement programs for sepsis much more embraced and much more effective.

Ryan Reynolds
That was Ron Daniels from the UK sepsis Trust. You also heard from Seung hoon gwon from Seoul National University in Korea to read Seung hoon's paper. Look out for a link in the show notes.

Unknown
Coming up. As more and more AI generated data gets out there, it could start causing problems for new models.

Right now, though, it's time for the research highlights with Dan Fox.

Dan Fox
Three huge star forming clouds in our cosmic neighborhood are actually all part of the same enormous structure.

Molecular clouds are dense regions of gas and dust where stars are born. Three such clouds, known as chameleon, musca and coal sac, are situated only about 6 quadrillion kilometres from Earth.

Researchers studied a 3d dust map of the region containing chameleon, musca and coal sac and found that these clouds arent isolated, but instead form a connected c shaped structure with a radius of about 50 parsecs.

Based on these findings, the authors proposed that a single supernova turned an existing dust cloud into the newly identified c shaped structure somewhere between 4 million and 10 million years ago.

You can read that research in full in astronomy and astrophysics.

Satellite observations suggest that an oil well blowout in Kazakhstan last year may be the largest accidental methane leak ever recorded.

Researchers used a constellation of methane detecting satellites to calculate the total emissions from the leak by analyzing methane concentration concentrations and the rate of gas flow over time.

They estimated that the well belched 131,000 tonnes of methane into the atmosphere over nearly seven months in 2023, ranking it second only to the Nord stream pipeline event, which leaked an estimated 478,000 tonnes of methane into the Baltic Sea in 2022, after the pipeline was deliberately breached.

Methane is the second biggest greenhouse gas, contributing to the rise in global temperatures. Before satellites were capable of mapping and measuring methane emissions, super emission events such as this could go unreported.

The oil company finally plugged the leak by injecting mud deep into the cavity.

You don't need a satellite network to find that research, it's in environmental science and technology letters.

Unknown
Next up, reporter Jeff Marsh has a story about how feeding AI generated data to AI's can make things go a bit off.

Unknown
The Nature podcast is a podcast produced by the scientific journal Nature.

It features interviews with top scientists and researchers discussing their latest discoveries and research findings across various scientific disciplines.

Unknown
Large language models that power chat, GPT, Bard, and Gemini, amongst others, are amazingly good at sounding human.

Unknown
Nature podcast provides a valuable resource for anyone interested in science and scientific research.

Unknown
And they've clearly got good taste in podcasts. But the reason they're so good at sounding human is because up until a few years ago, the data these models were trained on was exactly that.

Unknown
Human Earth is the third transmitter producing electromagnetic.

Unknown
Every book, every article, every conversation that exists online that it could get its hands on was used for training.

But what happens when this podcast is transcribed and posted on the Internet, along with that description of the nature podcast?

Unknown
For anyone interested in science and scientific research.

Unknown
With an ever growing body of AI generated text, data hungry generative models of the future could well be training themselves on datasets with less and less authentically human data, and more and more of their own outputs.

Unknown
Here are ten tips to improve your workflow.

Unknown
Since large language models became widely available and publicly available to individuals for almost no cost, stuff has changed quite significantly.

Unknown
This is Ilya Shmailov, who at the time of writing this recent nature paper was a junior research fellow at the.

Unknown
University of Oxford, and I was a member of Oxford's Applied and theoretical Machine Learning group with Professor Yaring Gall.

So language models are statistical beasts that you can show a lot of written language, and you can ask those language models to learn on a very high level statistics of all of this human produced language, and then learn to produce language that looks very similar to what they observe written down.

So the amazing part is you don't require any supervision. It's pretty much like show it a lot of examples of text, and then it's capable of learning those dependencies to produce text that appears ideally indistinguishable from the original text.

Unknown
But could their widespread success become their downfall?

Unknown
So now it's quite common to use chargebitty, Gemini and other large and capable models to help you with writing tasks. I hope this message finds you well.

I am writing to kindly request an update on the progress of the report.

It's totally normal to ask it for recipe suggestions, or asking it for suggestions of what to do in London. Here are some popular suggestions. Visit iconic landmarks and at the same time, it's relatively common to ask it to speak in some other language with you and or explain to you different phenomena.

So I think it's fair to say that we rely on large language models a lot more nowadays as sources of information.

And as a result, since we then use this information for something. Yes, we start chatting with our friends using this part of chat GPG. We submit essays university.

Ultimately it ends up online and then scraped by other people who then train models on top of this artificially generated data.

One afternoon, me and my brother, the second author in the paper, Zack, we were basically talking about whether we expect that training language models in the world tomorrow is going to be easier or harder.

You expect it to be easier since a lot more data is going to be available and this data is going to come out of good language models. So you would expect that it's easier to actually train on it, since you know that a language model is relatively good at approximating it. But it's also going to be hard, because suddenly it will get slightly more annoying to discover datasets which are representative of all of the data.

And then we quickly went and started trying to mathematically model this behavior, and we discovered that actually it gets significantly harder to train those models, even the.

Unknown
Simplest of models with a somewhat smaller budget than Google, Meta or OpenAI. Ilya and his colleagues developed a stripped down model, starting with a pre trained language model that speaks relatively well.

Then it was time to fine tune it.

Unknown
What we do then is we grab a data set, a Wikipedia dataset that contains Wikipedia articles, and then we take this model and we find what's called fine tune on this article. We basically say, this is our source of truth. Now, can you please learn the language that is similar to the language that you will observe in the Wikipedia article?

And for the model from generation zero, it is a pre trained model that is then fine tuned only on the original Wikipedia articles. So what we do then is we ask this model that lured the Wikipedia articles to produce new Wikipedia articles that it would have written based on the article descriptions.

And then once we've written those artificial Wikipedia like articles, we then use those articles to fine tune the pre trained model again. And we keep on repeating this process whereby we take data, we fine tune on top of this data, we generate new data, and then we use the new data as a fine tuning data set for the next model to come. And we keep on repeating this process. And what we show is that as you keep on repeating this process, the model quality degrades further and further and further, and the variance of the outputs that it's capable of producing is shrinking.

Unknown
To get a sense of what happens to Ilia's model as it fed on more and more of its own generated data, I asked him if he could prompt it to describe the Nature journal.

Unknown
Yes. So if you ask the original model, fine tune on a Wikipedia article, you'll discover that it discusses nature journal quite well. It says Nature Journal is a peer reviewed scientific journal published by the American association of of Science. Im not sure if this is true.

Unknown
Its not.

Unknown
The journal was established in 1993 as an independent publication and has since become one of the worlds leading science journals. Nature publishes more than 1000 articles on topics related to nature.

Unknown
Okay, so making some mistakes but sounding pretty human.

Unknown
But then if you keep on repeating this process, youll discover that the quality of the outputs will start degrading further and further. So for example, by iteration two, it starts repeating itself quite a lot. It says Nature Journal is a nonprofit organization. Nature Journal is a non profit organization. Nature Journal is a non profit organization. And so on.

Unknown
But listen to what happens when we get to generation nine.

Unknown
Nature Journal is one of those books where you can't help but feel like you're going to die if you don't do something about it. It's got something about it that makes me want to go back and read it again.

Model collapse is a degenerative process in learning from uncurated synthetic data. So think about it this way. You have a model that learned some concept and it learned it relatively well, but there is a little bit of error. It then goes and produces data that is used to train another model. But the new data already possesses this additional biases that were introduced by the model from generation zero. Then model at the generation one learns the data and also learns those additional biases. And since most of those biases came from the models that are pretty much trained in the same ways, models become more and more convinced that those errors are sources of data.

And what we show is that over time, as you get to generation n, it becomes overly convinced on the errors and a lot more iffy on the actual real data organization. Nature Journal is one of those books where you can't help but feel like you're going to die if you don't do something about it. It's got something about it that makes me want to go back and read it again.

In principle, you will always be experiencing this because it's more fundamental.

If we are using 50% real data, 50% non real data, the model collapse effects are probably going to be attenuated a little bit, and they're going to be probably not as prominent as if you used like 1090 split.

But at the same time, what we can be certain of is that model collapse will be happening.

Unknown
So does this mean that all AI generated data needs to be kept separate from data of a human origin?

Unknown
Yes and no.

What the paper talks about is uncurated use of synthetic data in training. That is to say that no additional filtering is done to separate away data that is explicitly erroneous or is definitely bad. And I'm certain there are ways to cheaply detect that some data is of lower quality than the other data, and we do not cover these cases. What we talk about is just an uncurated use of data, and it's an open question of how to train language models and other types of models on top of synthetic data to benefit from learning, yet also incur no additional degradation.

There is no answer to that question.

So maybe in the years to come we'll discover how to do this properly, but at the current stage, unfortunately, we don't really know what to do.

Unknown
That was Ilya Shmylov from the University of Oxford here in the UK talking to Geoff Marsh. For more on that story, check out the show notes for some links.

Ryan Reynolds
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. Emily, what have you been reading this week?

Unknown
So I found an article on nature.com that caught my attention, that was looking at how the psychedelic drug psilocybin causes lasting changes to pathways in the brain. So psilocybin is the active compound in magic mushrooms, and it's one of lots of psychedelic drugs that are being investigated as therapies for conditions such as depression and post traumatic stress disorder. And there's quite a lot of data that these compounds have a positive effect, sometimes lasting years after the treatment has ended. But researchers still don't fully understand the mechanism that underlies this.

Ryan Reynolds
So we've covered various psychedelics a few times on the podcast as well. And I. Yeah, it's really hard to work out how exactly they're having their effects on the brain. So what was the approach they used in this study?

Unknown
So, a lot of previous studies have looked at the effects on individual cell types or very specific areas of the brain. This took a much more holistic view of the brain. So they put seven participants in an fmri while they took psilocybin, and they imaged them before, during, and after they took this high dose. And they were able to get images of the flow of blood throughout the brain. And this is sort of a proxy to measure which groups of neurons are active and communicating with one another. And what they found was that psilocybin caused groups of neurons that normally fire at the same time to become desynchronized, so disrupting the usual pattern of neural communication. And the greatest effect was in the default mode network, which is thought to generate a person's sense of self, space, and time. So it's the part of your brain that is normally active when you're daydreaming.

Ryan Reynolds
Right. And when you say desynchronize, what does this exactly mean? Because it sounds like quite a dramatic change.

Unknown
Yeah.

Unknown
So desynchronization is stopping two areas of the brain communicating as much, and it does sound like a bad thing. It's disrupting the usual patterns of neural communication. But what this actually leads to is the brain becoming more malleable, more plastic. So it's been found in people with depression and other mental health disorders that they have these very rigid processes going on in their brain that are tricky to break. So being able to disrupt these pathways, it allows them the opportunity to rebuild them in a more healthy way.

Ryan Reynolds
I mean, this sounds like it could be very promising, but when we've covered psychedelics research on the podcast before, it's always been used in very sort of specific settings, right?

Unknown
Yes, absolutely. So this was all done in a very strict environment. They had doctors on hand in case someone experienced a bad trip. There's actually something quite interesting around that. They found less neural desynchronization when participants were engaged in an active task of audio and visual matching. So they were making their brain think about something else, and when they weren't involved in that task, they found far more neural desynchronization, which suggests, during a psychedelic experience, grounding. So bringing yourself back into yourself could be a way to get over a bad trip, to stop trying to look at external things, and to come back into your brain. It's also worth noting one of the participants of this study was actually one of the researchers.

They decided to take psilocybin, scan their own brain, and get the data from themselves, and reported on his experience. He said, I was inside the brain, and I was riding brainwaves. So his trip involved a brain itself, which I think is quite nice insight.

Ryan Reynolds
There, into a neuroscientist's state of mind, I think.

Unknown
Absolutely.

Ryan Reynolds
And so I wonder, what is the sort of next steps for this research we've discussed, like, how it could potentially be useful for certain conditions. So what is next to sort of, I guess, bring something like this to the clinic?

Unknown
Well, this data can't show precisely what is causing the potential therapeutic benefit but it does sort of offer clues. So it's possible that psilocybin is causing the brain network changes, or perhaps it's creating the psychedelic experience that causes parts of the brain to sort of behave differently. It will be good to untangle whether the blood flow directly correlates to the neural communication. As obviously with fMRI's, you're not measuring the actual electrical signals of the brain, but the blood flow to different areas. The researchers hope to conduct further experiments to investigate the effects of psilocybin on the brains of people with conditions such as depression.

Ryan Reynolds
Well, it certainly sounds like some fascinating research. I would have liked to have been a fly in the wall during some of those experiments, I think. But for my story this week, if you'll ride on my brain waves, I'm taking us up to the moon. Or at least, actually, I'm not taking us up to the moon, because this is a story about a moon mission being cancelled.

Unknown
Oh, what was this moon mission?

Ryan Reynolds
So, this was an article I was reading about in nature, about NASA's mission to go to the moon and go to the moon's south pole and map the ice and drill it there. And as we've talked about a lot of times in the podcast, a lot of countries, a lot of agencies are interested in the ice on the moon because it could be potentially used for fuel in the future, or it could be used for oxygen. And also, missions like this can just help us understand the moon better and the solar system more generally. And so the reason that this was cancelled was due to rising costs. And a big part of that was the volatiles investigating polar exploration rover. And this is a rover that's been assembled, but it seems that it will never touch down on the moon.

Unknown
So what are the reasons for this cancellation?

Ryan Reynolds
So it's mostly just delays. So there have been delays in building the rover and also in building the commercial lander, and that's pushed the launch date back all the way to 2025. And that's caused an estimated rise in cost of another $176 million. So NASA had an internal review after that, and they decided to discontinue this mission because they. To make sure they have enough budget for other missions that they're planning as well. And also, NASA said that it's a bit of an uncertain budget environment for them at the moment. And the reason for that is, in 2024, they had a smaller budget than they did in 2023. And the budget proposed for 2025 is only a 1% increase, which is actually lower than the rate of inflation. So that means it's a real terms cut.

Unknown
So this is not delayed, this is, it's done. It's not going to the moon at all?

Ryan Reynolds
No, it doesn't seem so. Well, maybe the rover will go, but we'll come to that. There's just been sort of spiralling costs, and that's been associated with delays in building the rover and also the commercial lander that would actually put it on the moon. So one thing to note here is the rover has been built by NASA, but the lander that would actually put it on the moon has been built by a separate company, this Astro botic technology.

Unknown
So is this rover doomed for the scrap heap?

Ryan Reynolds
No, there is a potential that if you want to, you could use it. So NASA is actually looking for partners to take over the rover because basically it's complete. So another agency could perhaps take over this mission if you've got, I don't know, I guess a few hundred million dollars.

Unknown
Spare one for the pocket change?

Ryan Reynolds
Yeah, maybe it's outside of our budget, but if there's any space agencies listening that are interested in and perhaps they can take it over. If no one does, though, they'll probably use the components of it for future lunar missions. But this has also been a bit of a surprise for researchers because, you know, it seems strange at this point when the rover is complete to basically scrap the mission altogether, because some researchers have asked, like, why don't they just put in storage and wait for a better moment to do it? But that's the decision NASA has made for now.

Unknown
So are there other projects looking at going to ice on the moon, or is that it for ice?

Ryan Reynolds
No, no, there are more missions. So we talked about on the podcast before, there are other nations that are interested in going to this part of the moon and looking for ice there, and NASA itself are not giving up on it. So despite what has happened to this mission, there's another mission scheduled for later this year, the polar resources ice mining experiment one or prime one, and that's going up on a lander built by a different company, intuitive machines. And they've successfully landed the spacecraft on the South Pole already. So, you know, they have a bit of a track record record here.

Unknown
Well, I'm sure we'll have more on moon missions in the future. It's a staple over here. Thanks, Nick. I think that's all we've got time for listeners. For more on those stories and for where you can sign up to the nature briefing to get more like them, check out the show notes for some links.

Ryan Reynolds
That's all for this week. As always, you can keep in touch with us on X. We're at Nature Podcast, or you can send an email to podcastature.com, and if.

Unknown
You fancy leaving us a review, you can do so wherever you get your podcast. I'm Emily Bates.

Ryan Reynolds
And I'm Nick Petra Chow. Thanks for listening.

Nature Podcast
The Nature podcast is supported by Nature Plus, a flexible monthly subscription that grants immediate online access to the science journal, Nature and over 50 other journals from the Nature portfolio.

More information at go dot nature.com.

Unknown
Plus, even when we're on a budget, we still deserve nice things. Quince is a place to scoop up stunning high end goods for 50% to 80% less than similar brands.

They have buttery soft cashmere sweater starting at $50, luxurious italian leather bags, and so much more.

Plus, Quince only works with factories that use safe, ethical, and responsible manufacturing. Get the high end goods you'll love without the high price tag with Quinn's. Go to Quinn's dot style for free shipping and 365 day returns.