← The Good Question Podcast
The Good Question Podcast
AI & Ancient DNA Prof. César de la Fuente On Discovering Next-Gen Antibiotics
00:23:01
Loading summary
Transcript58 lines
- [00:01]
A
You're listening to the Good Question Podcast with Richard Jacobs. Our goal is to make each of our guests exclaim, hmm, that's a good question. I don't know the answer. Because when that happens, it means you, the listener, may be inspired to learn more beyond the interview and to ask great questions yourself that lead to new insights. In this podcast, we cover historical and current anthropology, comparative religion and history. Welcome. And let's get started.
- [00:36]
B
Hello. This is Richard Jacobs with the Big Question podcast. My guest today is Cesar de la Fuente, or Cesar de la Fuente. We're going to talk about getting new antibiotics from looking at the DNA of Neanderthals, woolly mammoths, animal venom, and now proteins of ancient microbes. Very, very interesting stuff. Cesar's at professor at the University of Pennsylvania. He runs a group called the Machine Biology Group. He's one of the youngest tenured professors in the history of plant medicine. So very welcome. Cesar, thanks for coming.
- [01:05]
C
Thank you, Richard.
- [01:07]
B
Great to be here. Yeah. What is. Tell me about the Machine Biology group. What is that?
- [01:10]
C
Yeah, so that's basically, you know, it's the. The group that I direct at University of Pennsylvania, and it's this really transdisciplinary team that encompasses people with a wide range of expertise.
- [01:23]
B
Right.
- [01:24]
C
So, for example, we have experts in computer science, experts in chemistry, experts in microbiology, engineering, even physics, all working together in this really incred. And I'm really lucky to be able to work with people that come from all around the world, come to my lab, to this group to kind of to try to push the boundaries of science and the boundaries of knowledge. So it's really a fun, you know, I'm really lucky and fortunate to be able to work with such incredible people.
- [01:53]
B
Yeah. So how did you get the idea and. Well, first of all, why antibiotics? Why is it important that we pursue new ones? I don't know if people are aware of what's going on with current antibiotics.
- [02:03]
C
Yeah, there's a. There's a huge global health crisis, so that we call antimod Microbial resistance, or the acronym that we use is amr. And these are bacterial infections that are increasingly resistant to every drug that we have available. And those drugs, we call them antibiotics. So these infections are currently associated with about 5 million deaths per year around the globe. And if we don't do anything about it, if we don't come up with no solutions, by 2050, that number is projected to double to double to 10 million deaths per year, becoming the number, the number one cause of death in our society. Even, even surpassing cancer. And so, you know, we take that very seriously. I think it's one of the greatest existential threats to humanity. And so in my team, we're trying to, we're trying to think outside of the box to see how we can, you know, how we can come up with new molecules that might be able to target bacteria, and also how we might be able to find new molecules in places where perhaps people haven't looked before. Or how do we develop tools, computational tools that can enable us to accelerate the pace of discovery.
- [03:13]
B
So why look in the, the DNA of like ancient animals or, you know, in these far off places? They wouldn't be exposed normally to the, you know, I would think to the bacteria that, you know, that cause us problems. I mean, why not use organoids and why not try to again, deal with the, deal with the bacteria that are killing people, but in the lab environment or in a mouse model or things like that.
- [03:35]
C
Yeah, we, we also do all of the above. But so essentially a lot of the work that we have done, you can cluster it in two main areas. One, we're trying to mine the code of life for new functions, for new antimicrobial functions, for example. So we've developed methods that enable us to mine or explore what nature has produced.
- [03:57]
A
Right.
- [03:58]
C
We think of evolution as this planetary scale optimization process that we can learn a lot from it and we can extract a lot of useful molecules and useful information from it. So we've done that. And we've done that by mining extant biology, but also assess ancestral biology, like you mentioned, to try to find new antimicrobial molecules. We've been able to successfully find functional molecules in the human proteome, which is, you know, encoded in our, in our genetic code, but also in our closest ancestors, Neanderthals, Denisovans, and then we scale that up to all of ancient biology. And so we took this journey with our AI systems through the genetic data of organisms like ancient penguins that were extinct in the 50s, or magnolia trees that disappear throughout evolution, or even woolly mammoths, ancient elephants, giant sloths, and many, many other creatures that used to roam around our planet at one point, but unfortunately, through extinction events, they disappeared. But now we have this opportunity where we have access to some of their genetic data. And the way we think about this is we think of molecules as documents of evolutionary history, where we can extract, you know, potential therapeutics, but also learn more fundamentally about how evolution did things and how, and perhaps why evolution did certain. So that's one whole cluster. And the Other whole cluster is that in addition to learning from Mother Nature, we're also developing generative AI models that can create new to nature, completely artificial molecules that belong to the realm of the artificial.
- [05:36]
A
Right.
- [05:36]
C
They're molecules that evolution has never thought of. We think perhaps they give us a fighting chance at targeting antimicrobial resistance and perhaps other major problems facing humanity.
- [05:48]
B
Yeah, I guess there's a lot of ways to slice it. What about, I don't know, I know they wanted to make antimicrobial materials, maybe materials of compounds that bacteria would never normally see that have like a high, high roughness factor that, you know, like bacterial membranes would get shredded kind of, you know, getting close to them. I don't know, I guess it's all different angles.
- [06:08]
C
But yeah, I mean, I mean if you think about it, right, some of these, the contemporary pathogens that were, that were exposed to today, right. In the world, most likely they've never seen some of these molecules that existed, you know, many, many of years ago. So on the therapeutic side, you can think of this as a way of extracting potential therapies that can help us address present day problems like the problem of AMR by looking at the past, by using this ancient data essentially as relics or fossils that we can bring back, we can synthesize in the lab and we can, in some cases, some of these molecules that we've identified, they, they are no longer, as far as we know, expressed in nature today. So they, they really kind of in some cases disappear throughout evolution. And so we're bringing back those molecules to, in this case, to counter drug resistant infections.
- [07:02]
B
Yeah, but why, why would they be any more likely to fight an infection versus, I don't know, like, you know, the bark from a tree along the side of the road that grows, you know, like, I don't know, it just, it just seems odd. Like why, why would that be a better area to look at than anything we have today around?
- [07:18]
C
Yeah, so, so we' also explored extant biology, biological data. So we're doing all of it right. So we don't think it's necessarily better or worse. We just think it gives us, it helps us expand the combinatorial sequence space that we explore in evolution. And people had not really done it before at this scale, looking at both extant and ancient biology as a source of antimicrobial molecules. So we think from a fundamental perspective or a curiosity driven perspective, we think it's interesting because we can learn about how evolution did certain things and how, you know, certain mutations, you know, incurred modifications in the function of those molecules.
- [08:00]
B
So I understand you want to expand the combinatorial space. That makes total sense. What even if you do develop a new antibiotic though, I mean, I don't know from what I've heard is big pharma is really the only ones that have enough money to get it through clinical trials and they don't want it because it's like, you know, a 10 day dose and you're done. So who's going to, I mean, are there government grants for this stuff or how do you get this stuff to market even if you find it?
- [08:24]
C
Yeah, you're absolutely right. It's, it's extremely difficult. It's really hard to fundraise to, to actually develop molecules through ind. Enabling studies and through clinical trials. So you know, we, we've spun out a company to, in an attempt to try to do that, but we know it's going to be really tough like you mentioned. But nevertheless, the way we think about this is that ultimately it would be for the benefit of humanity.
- [08:47]
B
Right.
- [08:48]
C
We were discuss how these infections, they, they're a big problem, right. They're, you know, associated with 5 million deaths per year in the world. And you know, I think also importantly, and perhaps this has not been highlighted sufficiently is that without working antibiotics, modern medicine as we know it, it would really collapse.
- [09:09]
B
Right.
- [09:09]
C
Because you need them for a lot of routine interventions in hospitals, things like, you know, surgeries even, even like a C section organ transplantation. You need them for child, you need them to treat immunosuppressed patients, for example, cancer patients. And so really they are a cornerstone of a lot of the medical procedures that happen every single day in hospitals around the world. And so I think it is important to think of trying to develop new antibiotics as a sort of like a benefit to humanity as opposed to trying to make profit. Even though of course, are there any
- [09:46]
B
that have been made that are just sitting there with lack of funding?
- [09:50]
C
Yeah, so there are instances of antibiotics that have been approved. But then, you know, hospitals do not have the incentive to use them right away. They only use them as a last resort situation. And so it's hard for companies to, to kind of, to obtain a profit.
- [10:07]
B
Right.
- [10:07]
C
So some of the, some of the ideas that people have thrown around is perhaps getting government subsidies involved or nonprofits that perhaps could do something about this and help fund some of the, some of this stuff to, to help ensure that companies are don't, you know, are successful ultimately.
- [10:28]
B
Right.
- [10:28]
C
And to be able to incentivize other companies joining this area. The other thing that maybe you find interesting is that we're also experiencing a decrease in, in the talent pool because there is less and less funding to do this kind of work in AMR. You know, we're getting less PhD students trained and you know, there's, there's sort of like a brain drain that we have experienced over the last decade or so. And so that's also concerning.
- [10:54]
B
Right?
- [10:54]
C
We, because we need to, we need to train the next gener to, to help, help us sort of help us combat amr.
- [11:02]
B
What about phages? I mean bacteria, they have to still have phages that kill them, otherwise they would spread all over the place. How come no one looks at that or at the protein coat of, you know, various bacteria that are AMR to see what's changed? Like why not a more direct focus?
- [11:19]
C
Yeah, that's, that's a great idea. There are, there are a number of groups around the world doing very incredible work in a phages. So just for your audience, bacteriophages, these are viruses that only infect bacteria so they don't really do anything against human cells. And people have attempted to leverage phages as a way of treating drug resistant infections. And in fact, if you, if you study their history, they have this really interesting history where they were using the Soviet Union. Even today, if you go to the Republic of Georgia, you can buy them at pharmacies in little cocktails, you know, to treat, to treat infections. And so I think that's a really valid alternative to conventional antibiotics. I, the more ways and the more different approaches that we, we try to develop to counter amr, the better for the world. So I think those are all great, great avenues to, to, to pursue.
- [12:11]
B
Are there other commonalities? If you look at the posterior membrane of, you know, AMR bacteria, so in
- [12:18]
C
some, in some cases the membrane can change or you know, bacteria that become resistant, they can do that in a number of ways. One of them is by changing their membrane. You can think of it as they, they change it, shield themselves more effectively from the outside world and from antimicrobials. But they can also, you know, they can also degrade antimicrobials. They have these efflux pumps that enable them to pump out antimicrobial molecules. So they have a huge arsenal of mechanisms that enable them to survive essentially. Right. Let's not forget that bacteria are some of the most ancient organisms to have ever existed it on earth. And so they have a huge toolkit of little tricks that enable them to essentially survive and adapt Right. And so, yeah, I'm no longer surprised by the incredible sort of the clever tricks that they have up their sleeve to kind of overcome any situation where
- [13:19]
B
else they use efflux pumps. I mean, what else can they use? What are some of the things you've seen?
- [13:23]
C
Yeah, they have enzymes that enable them to. So when they come across an antibiotic molecule, they can degrade it or break it down. They also form biofilms. That's an adaptive response where they, they form these multicellular communities of bacteria. So they, they form these little houses, these little sort of towns and cities, and they, what they do is they encapsulate themselves in this, what is called a matrix, which you can think of it as a really effective shield all around the whole microbial city. And that matrix is composed of things like sugar and DNA and proteins. And so conventional antibiotics cannot really pene penetrate through that really thick and effective shield. And so that's another way that they can do it. You know, they, and, and you know, we can, we can talk for a lot longer, but they just have all, you know, a wide range of, of survival strategies, essentially.
- [14:16]
B
I mean, there's certain bacterial genus that tend to form these defenses more effectively than others. You know, gram negative, gram positive and
- [14:25]
C
gram negative, gram negative bacteria. So you can classify bacteria like you mentioned in gram negative and gram positive, gram positive bacteria only have one MEMBR thick membrane, and gram negative bacteria have two membranes. They have the outer membrane and then they have the inner or cytoplasmic membrane. So the bacteria that are most problematic in hospitals that kill the most people are gram negative pathogens. And if you look over the list of, from the World Health Organization, the usual suspects that you'll find there as the pathogens most dangerous to our society, those are primarily gram negative, and that's because they have two membranes. So it's really hard to kind of panic. Penetrate into them with molecules. And yeah, so they include names like Pseudomonas aeruginosa or Acinetobacter baumani. Almost unpronounceable names, but that are essentially these really clever pathogens that can cause a lot of harm in humans.
- [15:23]
B
Are there any bacteria that are really good at competing or outcompeting other bacteria, regardless of whether they become antibiotic resistant or not?
- [15:33]
C
Absolutely, yeah. Bacteria. So essentially, in any environment, bacteria, the way they survive is by colonizing a particular environment. And then in the process of doing that and in the process of growing their colonies, they have to outcompete surrounding microbes. And so they are really Experts, you know, a lot of them are really experts at doing that. For example, we've studied a bit in detail Pseudomonas aeruginosa, which is a gram negative pathogen that is highly effective at out competing surrounding microbes. But there are many, many others.
- [16:05]
B
But are there any ones that are harmless to people even if temporari? We have a whole ton of them in us, but they're very good at out competing, you know, bacteria we don't want.
- [16:14]
C
Yeah, that's absolutely true. So there are. So the vast majority of microbes that exist in the world, they're actually beneficial, they're not pathogenic. It just happens that the pathogenic ones, because they kill humans, they get a bad reputation. But in Anonas, we have millions of microbes that live in our mouth, on our skin, in our urinary tract, in our intestine that do a lot of good things for us. Actually they have co evolved in many cases with us and they help us digest our food, they help us fight off disease, they out compete pathogens and so they, they do all these really good things for us. And, and a lot of them are really like you mentioned, really experts at, at fighting off actually pathogenic bacteria that would be bad for, for us, for the host.
- [17:00]
B
Again, is there any, any commonality in how they out compete? I don't know. Are there like so what happens with, with these anti, antimicrobial resistant bacteria, you throw all the, you throw all the antibiotics at them are by them. Are they hurt by them? But not as much. Are they using the antibiotics as somehow raw material to even grow faster and stronger? Like what's, what kind of dynamics do you see?
- [17:23]
C
So yeah, it's a combination of all of the above and it really depends on the microbe. So in some cases they, they over express these efflux pumps that enable them to pump out, you know, the toxic microbials that we throw at them that are toxic for them.
- [17:37]
B
Right.
- [17:38]
C
In some cases they form biofilms so the antibodies can no longer penetrate into and reach those microbes to kill. In some cases, you know, they produce little molecules that operate like a form of chemical warfare to kind of kill surrounding microbes. And so there's all these, all these little, all these different tricks, right? And the way that resistance, one of the ways that resistance can spread in a microbial population is in plasmids. And these are circular pieces of DNA that can be easily transmitted between microbes and that they confer resistance, you know, through different mechanisms. But is this really cool way that Microbes, microbes have of sharing genetic information in this circular DNA structures to kind of ensure the survival of the whole community, of the whole microbial community.
- [18:26]
B
Any promising compounds that you found that you can talk about, you know, the mechanism of action, whatever you can say?
- [18:33]
C
Yeah, absolutely. So in our whole efforts to mine the code of life, so that means, you know, genomes, proteomes, metagenomes. With machine learning models, we've been able to identify, for example, Mammuthosin 2, which is a molecule encoded in the woolly mammoth proteome that is quite effective even in mouse models of infection, at reducing infections. We found Elephacin 2 from the ancient elephant and really a lot more. So in summary, a lot of the work that we've done has enabled us to discover millions of potential sequences with antimicrobial properties by really mining biology at scale. And out of all those, we have probably by now synthesize thousands of them using chemical methods. So we have these chemical robots that enable us to synthesize whatever we design or discover on the computer. And then we've probably tested and validated in preclinical animal models, hundreds of them. So really we're doing this at a pretty large scale.
- [19:42]
B
What do you think will be possible in the next year or so? The AI models got stronger to where you compare many more millions of compounds. Now what do you think is going to happen?
- [19:52]
C
So what's already happened is that the AI models are getting really good, really reliable, really accurate at predicting antimicrobial function. And this is because we're training them with really high quality standardized experimental data. And this is a hypothesis that we proposed a number of years ago. We call it the standardized data set hypothesis, where we predicted that if we trained our models with really good quality data, their performance would scale with the amount of data that we fed them. By performance, I mean simply the accuracy of those models. And so when we, we started to track these metrics, I think it was in 2021, the hit rate was around 60%. And by hit rate, I mean when we actually make those molecules through synthesis in the lab, how many are actually active against bacteria in ground truth experiments? Well, that accuracy right now is at 95 plus percent. And that's because our models have gotten increasingly accurate and reliable. And, you know, so we've basically almost solved that problem ob a very specialized narrow area or for a very specialized objective function, which was to predict whether a particular molecule will be antimicrobial or not. But now that we've reached 95 plus percent accuracy. I think we can, you know, we're doing a lot less experiments than we used to, right. Like, and I think that's quite telling of what some of these AI tools can do that if you train them really, really well with high quality data, they can become incredibly good at really specialized fields. And so what we're doing now is we're extrapolating that and we're creating high quality data sets for many, many other parameters, including toxicity, stability, folding PKPD profiles, all these other parameters that you actually need to make a drug. And we're trying to train models in each of those areas to then have them achieve 95 plus percent accuracy. And in parallel, we're also working on AI agents that will be able to tap into each of those models needed to give a really, you know, a really high accurate response to the user.
- [22:06]
B
Where can people find out more about your work and follow up? What's a good, like, centralized place?
- [22:10]
C
Probably they can check out our website. We also post on social media so we can share that information on this podcast.
- [22:18]
B
Okay, very good. Thank you so much for coming on the podcast. I appreciate it.
- [22:22]
C
Yeah, thank you so much for having me. If you like this podcast, please click the link in the description to subscribe and review us on our itunes.
- [22:31]
A
Thank you for listening to the Good Question podcast. Please email supporthegoodquestionpodcast.com if you have any referrals to great guests for us to interview, visit thegoodquestionpodcast.com to hear more interviews. And please help us spread the word by rating and reviewing us on Apple podcasts, iTunes, Spotify, YouTube, or wherever you listen to this podcast.