Does every scientist need an AI co-scientist? How two professors solved a years-long viral mystery — Podovirus podcast
1 hr 09 min

Does every scientist need an AI co-scientist? How two professors solved a years-long viral mystery

The conversation

Episode notes

Imperial College London professors José Penadés and Tiago Costa explain how Google’s AI co-scientist proposed a key idea behind their already completed, unpublished research on phage satellites. Listen to learn how these genetic elements borrow different phage tails to reach different bacteria, and where the researchers found AI suggestions useful, incomplete or wrong.

The researchers found that DNA-filled satellite capsids can acquire free phage tails outside cells, with adapter and connector proteins controlling which tails they can accept.

The AI proposed testing different tails and examining adapter and connector proteins, but omitted intermediate mechanisms and prompted no changes to the completed experimental manuscript.

The guests suggest earlier AI input could have reduced fruitless experiments, while their separate plasmid example shows how prevailing ideas in the literature can produce plausible but wrong hypotheses.

Episode timeline

  • 0:00 Introduction: Testing AI against unpublished research
  • 2:12 Phage satellites that build their own capsids
  • 10:13 From a conjugation question to an AI test case
  • 24:35 Patent secrecy and the limits of the AI’s explanation
  • 46:29 Which experiments could earlier AI input have saved?
  • 1:05:03 A plasmid question exposes the AI’s literature bias

Learn more

Studies

Guests and background

Transcript

AI-transcribed and edited for readability. May contain errors or omissions; check the recording for exact wording.

Select a word or timestamp to listen. Use arrow keys to move between words.

  1. Jessica Sacher

    Hello, welcome everyone to the Podovirus Podcast. So today we have a special episode, and I might say that every time, but we're doing lots of different things all the time these days. And I heard about this exciting AI story that's also a phage PICI story, so it was the obvious choice of topic. I'm so glad to have convinced José Penadés and Tiago Costa, who are professors at Imperial College London.

    They have been working together, I'm sure we'll hear a lot more, but they have been working in this space of mobile genetic elements, phages, specifically phage-inducible chromosomal islands or PICIs. And they were working on this, and apparently Google let them use its unreleased co-scientist tool, which is an AI tool. And they were able to use it to, it sounds like, rederive

    a body of work that they had not yet published, but had been working on for a couple years. They gave the AI tool this research question and a little bit of background, but none of their unpublished data. The co-scientist came up with the same hypothesis that they had recently proven, but not yet published. So this is a story that we wanna dig into, and you might have seen a bunch of coverage of this lately.

    It's been on, like Forbes, The Economist, BBC, so they made it to the mainstream, which doesn't happen all the time with phages. So, I'm sure, yeah, you'll be able to look into that coverage too. But we wanted to talk to our scientists, our phage scientists especially, and to everyone who might be curious about what these AI tools useful for, and how do they actually start using them?

    So I'm very excited to have you both here, and of course, my co-host Joe Campbell is here too. So we'll get deep in and hopefully find out more. What are the pros and cons of using this kind of tool, and what are the limits right now? To start, tell us just a little bit about what you were studying when you came across this tool? How did that even come into your lap?

  2. José Penadés

    Okay, for more than twenty, twenty-five years, we have been working with these PICIs. They're phage satellites, so in theory they use other phages for induction, for mobility, and at the beginning, we had quite a few issues because people just thought that these PICIs are defective phages. Okay, for probably for more than ten years, we were fighting just to put the PICIs as a new family of mobile genetic elements, a family of phage satellites.

    And probably, I don't know, ten years ago or twelve years ago, we discovered a peculiar PICI family, the capsid-forming PICIs, or what we call cf-PICIs. They have all the genes required to produce the capsid, the small capsids, and to package the DNA into the capsids, but they still require tails for creating effective particles.

    This is quite unusual because the majority, all the satellites, the other satellites, just hijack probably everything from the phage. Okay, but this is a very peculiar, I think so. And we realized that these capsid-forming PICIs were present as exactly the same DNA, the same element in many different species. Sometimes, we put as an example one cf-PICI that was in seven species from five different genera.

    It was a bit surprising. We thought this was probably a new mechanism of gene transfer, allowing these elements to go between species and even between genera. And this is what we have been doing in the lab for a few years to try to understand the mechanism.

  3. Jessica Sacher

    Exciting and then were you also working with Tiago's lab, or how does that come into play?

  4. José Penadés

    So, Tiago is a structural biologist, probably he will, he can introduce himself later. It's quite funny because, he works in another important mechanism of gene transfer, he's an expert on conjugation. We really wanted to know what these capsid-forming PICIs look like structurally. They have small capsids that can package PICI DNA.

    Because theoretically, the genes involved in the production of these small capsids, the capsid-forming PICIs, the capsids are related to phage proteins, but somehow they have evolved the ability to produce these small capsids and just to package the PICI DNA inside. And so, we started working with Tiago because he is the expert in getting, all these kind of structures and things.

    And in fact, he solved the mystery, but I think it's better if he introduces himself to explain.

  5. Jessica Sacher

    Go ahead.

  6. Tiago Costa

    Yeah, well, we we solved a a a a fraction of the of the mystery. So we we still don't know the the f the full story, but, uh, I think we are on the right track. The AI co-scientist’s results also seem to agree with our experimental data and hypotheses.

    Yeah, so I come from a different mechanism of gene transfer from conjugation. But, my lab and José's lab share an interest in how bacteria, either directly transfer the DNA among themselves by conjugation, or how they use the phages to mobilize the DNA.

    José and I teamed up to study capsid-forming PICIs from different angles. My lab's role was to understand at the molecular level, how those cf-PICIs are formed structurally, how these capsids are able to, you know, take, one protein, the phage protein, and change it, to make this structurally, very different, entities.

    So just to go a little bit more in detail. We understood that there are some insertions of protein sequences into the capsid forming proteins. We think they determine the different symmetry of the capsid-forming units and make them smaller, so that they can only accommodate the DNA of the cf-PICIs and not the DNA of the site.

    Which sort of makes it smaller so they can hold only the DNA of the PICI.

  7. José Penadés

    We should say that the PICI genome is usually about a third the size of the helper phage genome. The PICI is around ten to fifteen kilobases, and the helper phage is around forty-five. So a classical mechanism of phage interference is for the satellites to produce small capsids where just the PICI or the satellite DNA can be packaged. Okay,

  8. Jessica Sacher

    hm got it.

  9. José Penadés

    These cf-PICIs produce small capsids. They also have a specific terminase to package the PICI DNA, not the phage DNA. It's a specific system for packaging the cf-PICI DNA into the small capsids that are formed by proteins encoded by the satellite.

    That was another, unique thing compared to the satellite. Okay,

  10. Jessica Sacher

    And I think I saw in one of your papers that these are not considered parasites because they're not having a detrimental effect on the phage. Is that right? Hm,

  11. José Penadés

    yeah, So for many years, we I think it probably was in terms of selling, the papers. It was nice to tell the war between the parasite, you know, and. But in this case, the prototypical member of this family, we didn't see an interference with phage reproduction. We don't know whether that's true for other cf-PICIs.

    We published a paper about two years ago showing that some phages might suffer from inducing PICIs because some of the PICIs have anti phage systems. So even if you have some cost inducing these PICIs and if you can move them against other phages, you can also benefit. So we don't see now them as a parasitic relationship, might be so, but we might see more, we see them more as a yeah, it can be more synergistic, or, you know, depending on the scenario. They are more friends than we thought, probably fifteen or twenty years ago. Okay,

  12. Jessica Sacher

    What made you want to use this AI co-scientist tool? Where were you when you figured out that was an option? And why did you set it to work on this problem?

  13. José Penadés

    Tiago started using it for his work. Tiago, go ahead.

  14. Tiago Costa

    The Fleming Initiative at the college put us in contact with Google. They knew that Google was developing this new AI model tailored for scientists. And they wanted, to engage with scientists, so that they could understand how good the LLM was.

    Initially, this wasn't related to PICIs. My lab was trying to answer a question about conjugation that had been open for seventy years. So we were interested to understand, how conjugation is actually initiated, okay, and we wanted to understand the molecular determinants behind the T equals zero of conjugation.

    So we have challenged the algorithm, and the algorithm came up with several hypotheses, five different hypotheses, And we realized it would take us many months to understand whether the highest-ranked hypotheses would be correct. José was in that meeting and immediately saw an opportunity there. So why don't we challenge the algorithm with the experimental data we already have in our hands? I don't know if we were about to submit or had already submitted to one of the top-tier journals. Basically, we turned upside down the strategy of Google instead of having AI-driven hypotheses that would be validated a few months down the line.

    We already knew the mechanism behind the spread of cf-PICIs across different bacterial species. We challenged the algorithm with that exact same question, and it was astonishing. I mean, the the one of the the the top or the top hypothesis was basically a recapitulation of what we have, uh, observed, uh, in the in the lab.

  15. Jessica Sacher

    Wow. So were you surprised? Yeah, go ahead, José.

  16. José Penadés

    He didn't say this, but he was really impressed by the hypothesis the co-scientist suggested for conjugation. I'm not an expert there. That's why I said, “Okay, this looks really good.”

  17. Jessica Sacher

    oh yeah, hm, hm,

  18. José Penadés

    I need to. I need to take some advantage here. Because if the hypothesis provided by the co scientists for conjugation were correct, probably I wouldn't have been interested in it, but he was really impressed. I said, Okay, this is looking very good. So we changed the approach, and, it was quite impressive. The audience would probably like to know how this cf-PICI is mobilized between very different species is very simple. Okay, they can produce these capsids with the PICI DNA inside, okay, and that was a mistake we did for many years.

    We always thought that whatever is released after phage infection or whatever, it's an infective particle.

  19. Jessica Sacher

    yeah,

  20. José Penadés

    Okay, what we realized in this paper is that these PICIs can be induced even without induction. As soon as the cells lyse, what is released is just the capsid PICI with the packaged DNA. Annotate the cell. They have the ability to take tails from different phages, free tails has to be free tails, okay, and then, depending on the tail, that will determine, because we know for many years that the tails in phage biology determine the tropism. Okay, so that's it, okay, so there were many things on the paper, okay, until we realized that it's induction or non-induction, the cells lyse. Okay, the capsid, the packaged DNA will be released, no tails, and then outside will be free tails. We knew for many years as well that there were many tails that are producing excess, okay, and

    even, during the purification, they already showed that there were more tails than the infective particles or whatever. And then it's for the same capsid, depending on the tail that will bind there, that will determine the tropism. Okay, so, and then the one of the hypotheses that Google says is like, you need to check the possibility that these caps can interact with different tails from different phages.

    Okay, the system didn't talk about how that happened or that was induced, so it was just the final picture. Many steps in the middle and many new concepts in the paper are missing from the Google document, but the final result was right, and it was also quite impressive. We knew from the genetics, and now Tiago is solving the structure. Hopefully will be soon that there were two key players in this. The ability of a capsid forming or binding to one tail or another is two proteins called the adapter and the connector. So if you swap the adapters and the connectors, you always swap the ability of a capsid forming to bind to some tails or others.

    Okay, the co scientist also suggested we should look on the adapter and the connector.

  21. Jessica Sacher

    wow.

  22. José Penadés

    Okay, so it was so many things that were not, you know, in the output from the co scientists, but the key one, the tails, that was there, okay,

  23. Jessica Sacher

    wow,

  24. José Penadés

    and it was another one. So we put this in, you know, one interesting thing was that the system also suggested checking conjugation as a new mechanism, because conjugation is more promiscuous than, transduction, theoretically, because of the tail. The narrow for a specific tail, okay, and we have been, we have started working on that as well to see if that's true or not. But that was quite interesting, hypothesis as well, okay, that this system provides.

  25. Jessica Sacher

    Tiago must have been excited that it chose your favorite area.

  26. Tiago Costa

    Well, it was back to the beginning, right? The first thing we tried was to understand

  27. José Penadés

    as a structural biologist, Tiago loves the adapter and the connector.

  28. Jessica Sacher

    Ah, hm,

  29. José Penadés

    Have been narrow or wide, we, you just need to mention the adapter because it's it's pretty cool that some adapters and connectors bind to one specific tail, you know, look to be very narrow, and others have the ability to bind to several tails, okay, from different species. And he's all the time, very excited to see how mechanistically, probably one or two residues, you know, he's. That's the word you need to mention to Tiago: connector and adapter, that's it.

  30. Jessica Sacher

    I love it. hm

  31. Joe Campbell

    So, to get a little better understanding of the process. So I assume, because you hadn't done this when you started the experiments, were the hypotheses generated by the computer things that you had thought about, and therefore you designed experiments to test them, or did you have ideas? I mean, how were the experiments designed by you that ended up leading to the conclusions that agreed with the AI generated ideas?

  32. José Penadés

    I agree with that because the timing is very important here.

  33. Tiago Costa

    Yeah.

  34. José Penadés

    So we have been thinking a lot, okay, how this machine got the right answer? Okay, and it's very frustrating; it's very frustrating because we have the answer there and we didn't see it. Okay, as a phage biologist, we know that the tail determines the tropism. Is that right? Yeah, we know that many years, okay,

  35. Joe Campbell

    definitely, yeah,

  36. Tiago Costa

    yeah,

  37. Joe Campbell

    yeah,

  38. José Penadés

    for years. So if one DNA using a tail is in many different species, it's because maybe it's used different tails, you know? But we were biased.

    That's the problem we had, and I'm killing myself for that because for many years, I always thought, and all people in phage biology will think that after infection, what you have are infective particles

  39. Tiago Costa

    but yeah,

  40. Jessica Sacher

    yeah,

  41. José Penadés

    with the capsid and the tail. And we knew something very strange. We have PICIs that can be induced, but we didn't get transfer. We couldn't understand this thing because we were wrong in the way we saw this thing. We were so biased. Okay, we always thought that after induction or whatever, it's released from a bacterium are infective particles, both for phages and both for satellites. Okay, and I think that's the big thing. So Google didn't understand what's happening here, but make the most important connection. If you ha, you are in many different species, and to go to these species, you need a tail, you are binding to different tails; that's it.

  42. Joe Campbell

    Well, if I could just go back to the order. Am I hearing you say that you had data and you weren't exactly sure how to interpret it, and then you realized that this was telling you how to interpret it?

  43. José Penadés

    No, we had the data for years.

  44. Joe Campbell

    Yeah,

  45. José Penadés

    Okay, so, for example, we have the PICI, we use the capsid for MIS from Klebsiella, okay, and we induce the PICI. So this is a prophage there that is a helper prophage. Until the paper will be published. For us, a helper prophage means a phage that will provide whatever is required to produce an infective particle, capsid and tails, or just tails. Okay, so we have a strain in Klebsiella one we induce, we show very nice, capsid forming replication, very nice packaging, no transfer.

  46. Jessica Sacher

    hm,

  47. José Penadés

    We thought this was because the recipients we used didn't have the receptor, but we were thinking all the time that they were infective particles. Okay, and the same happened to us in E. coli. And then one day, suddenly, talking to my student, we realized that in E. coli, there were several prophages. But when we induced them, we knew that the island was heavily induced, there's a lot of replication and a lot of packaging.

    Because we can purify the capsids from the lysate, but the transfer was very low. And then he was making mutants to delete. It was, I think, six prophages, one by one, to see what's the inducer or whatever. And then he realized that it was one day that he removed one, eliminate the transfer, you know, even the low transfer, but the island was still induced. And at some point, we thought that maybe we need phage to induce and phage to provide the tails, okay, and then we start connecting the dots. And it has been known for fifty years or sixty years that you can get lambda or eighty lysate, mutant capsids, mutant in tails.

    You induce them, you take the lysate, you mix, and then you have infective particles.

  48. Jessica Sacher

    hm.

  49. José Penadés

    But the people that did these experiments, they never thought taking capsid from one phage and tails from another, never. Because we all thought that everything, you know, that that's the big bias, okay, so then that day, everything started making sense.

  50. Tiago Costa

    If you think I was excited when we understood that the adapter and the connector are the key factors here. You should have seen José on that day when he cracked this hard nut. He was very excited, cause he knew that this was something big, something novel.

  51. Jessica Sacher

    And you did this before the AI came into play. You cracked this, but you were spending a couple of years trying to get there.

  52. José Penadés

    More than a couple.

  53. Jessica Sacher

    Many years, a scary number. The AI came up with that because it didn’t have the bias that once you release a particle, of course, it has its tail and whatever it needs to be infectious. It was unbiased in that thinking. And of course, why not? Could you have this later meeting of capsid and tail outside the cell? And that's sort of what happened.

  54. Tiago Costa

    Yes. Sorry, José. I think we also need to clarify the chronology here, because there was another angle in play. Obviously, we saw an opportunity to patent these new elements because these elements are broad range, which is one of the limitations on phage therapy. You have all these narrow range these phages, using the phage therapy, they either only hit a bacterial species, or sometimes, strains within that species. And with this, with these chimeric particles, we met you tailor made and open up the you know, you have a tool kit where you can customize the target bacteria that you want to hit by for therapy or to make a diagnostic tool.

    So we kept this very secretive because we could only speak about this in public once the patent was filed.

    So it wasn't in the public domain. That's why we are sure that the AI system would never have had access to this manuscript or these ideas, because they were kept in a safe box in our computers.

  55. Jessica Sacher

    yeah,

  56. José Penadés

    Also, we challenge other systems as well with the same input, okay, and no, none of the other systems, you know, even more related, even to the co scientists, you know, and even so. I think the important thing we didn't mention this thing in the preprint. When we got the first round with the five hypotheses, we asked because, as I said, it's not really a real mechanism in terms of the capsids will be released outside the cells and they will, you know, was just okay. Check if this can bind to different cells without telling you how this interaction will happen, or when will happen, or whatever, okay,

    like, make a second round. And I said, Yeah, but this is, you know, challenging the system, saying, yeah, but this doesn't make any sense because how these particles are produced will be whatever. And the reply was crap, you know what I mean? So the system doesn't really understand what it's proposing, it's just connecting the dots in a way that makes sense, you know, just taking this simple thing that it details, determine the tropism. Maybe it's because it's using different cells, that's it, and it was there, we all knew that. Okay, so this is the type of things that for the people when they read the paper, they will think every single person will understand very quickly because it's obvious. And you know, but somebody needs to tell you. You know what I mean, It was frustrating because we spent quite a few years trying to understand it.

  57. Jessica Sacher

    yeah,

  58. José Penadés

    We have a strange and a lot of mutants, and we couldn't understand why we didn't get transfer. Is because just this tailless, we said this is a new entity, we call the tailless capsid forming, you know, the packaged DNA. These are the entities that are released. Okay, and

  59. Jessica Sacher

    yeah. Can phages, regular lytic phages, have this sort of mix and match of tail and capsid, or have you started to look?

  60. José Penadés

    Yes, I see that.

  61. Jessica Sacher

    That's so cool, do they?

  62. José Penadés

    we, if I need to bet, I would say yes, because it's same, you know,

  63. Tiago Costa

    yeah,

  64. Jessica Sacher

    Yeah, do they all have this?

    Yeah, they have these adapter connectors, like, do many phages have those? Is that conserved? Interesting?

  65. José Penadés

    Sorry, I think the biggest difference is that when I think, when a phage infects, the capsids and the compatible tail are already there. So we think that the number of capsids that will be released probably is lower. Okay, but the structurally, the capsid forming the proteins involved are exactly the same that you have in the typical H K ninety seven. You know, and this is the things that Tiago.

    But also we think that even other myoviruses or whatever, they work in kind of similar thing. The funny thing that we are working now, and you know, probably the people that it's good that we will say that in public, because then if somebody copies, can always get the credit. So you know now that there are many anti-tail, for example, anti-phage systems that block the tail formation, so when that happens, you will have capsid.

  66. Jessica Sacher

    oh, it's freely coming out.

  67. José Penadés

    So we really think the same will happen with phages, okay, and the phages will have the ability to inject DNA in different species by swapping tails. If these phages in the new species survive or not, this is another question, okay? We also show that the capsid forms when they go to another species. They have mechanism to hijack all the phages that are in the new species to also be mobilized and to be transferred.

    But mechanistically, we think, that will be the case, okay,

  68. Jessica Sacher

    Hm, yeah. So tropism is sort of like, you're like narrowing, kind of getting more high resolution on what it's actually defined by. Like, not necessarily just the tail, but specifically this connector point, maybe a little bit between.

  69. Tiago Costa

    Yeah, I mean the phenotype will be always determined by the deep of the tail, right? Whether it binds the surface or the receptor in the target cell. But it's true that now the neck region, where the portal, the adapter and the connectors are, will have a say in determining that tropism is not determined by direct binding to the surface or the receptor, but by generating a promiscuous or a very specific structure that binds only one or several phage tails.

  70. Jessica Sacher

    So cool.

  71. Tiago Costa

    And that's what we are looking after. Well, we sort of start to understand now what is imprinted in those structures of these proteins that determine that specificity or that promiscuity in binding only one type of tail or a set of tails.

  72. Jessica Sacher

    Wow,

  73. José Penadés

    funny. This, yeah, it, this capsid forming. So we proposed in the paper that we published in twenty twenty three that became the genes came from the HK97 phage. Let's say, you know, but they have a bulb, so this is not crossed all between the two proteins, even they are very similar in terms of sequence. Okay, so you have the capsid, the portal, the terminases, you know, the protease are very specific.

    Okay, so everything is very specific until they arrive to a point that is the connector and adapter. Each connector and adapter needs to bind to the capsid of the cf-PICI or the phage PICI, and then they need to bind to the same tail. So it will be quite a funny thing. You evolve something to be completely separate in terms of capsid forming things, and you know, the proteins involved that you cannot cross talk or you cannot, you know.

    But then the tail is the same, and probably will be a competition between them as well. So it's quite interesting to know what, what's happening there, you know? Because the idea we have is that probably we can create synthetic PICIs with the ability to bind to multiple tails. And if that's possible, we could solve one of the limitations for phage therapy. You know, We can create, we can deliver DNA to multiple strains or multiple species. Or, you know, this is the patent that Tiago was mentioning before.

  74. Jessica Sacher

    Yeah, yeah, I think this is so cool. It feels like one of the first fresh sort of phage host range discoveries that I can remember for a while. Like, it's another level of looking at the whole thing. And pairing it with the AI aspect, it feels like we have a set of tools we're all going to use.

    It's gonna be more, not less, like we're all gonna probably start using these tools. And now we can have someone with a beginner's mind, like who doesn't have all the assumptions of our hundred years of history. And then now we can get their input into just physically what would be probable or possible, and now we can kinda bounce off that whole other set of knowledge, built differently than ours.

  75. José Penadés

    I agree. For me, something that is kind of revenge, you know, no, nice for me. So I normally don't say these things, because people said that this is not the right things to say. But I think I succeed. Or some of the things we found in the lab were because I don't read too much.

  76. Jessica Sacher

    Yeah,

  77. José Penadés

    Okay, so I'm not biased.

  78. Jessica Sacher

    I love that.

  79. José Penadés

    Yeah, to some extent.

  80. Jessica Sacher

    You can't admit it. It's hard to admit.

  81. José Penadés

    I mean, you know, and so many times, you have a very good idea in the lab. You say, Wow, this looks so good. And then you go to Med and say, Yeah, somebody in nineteen eighty five thought the same, and then you have a huge admiration for these people. You know that you, but I wasn't biased, you know, that happened, for example, with the lateral and these kind of things, you know, because I was a completely ignorant, you know? I always said to my students, I trust your results because the students sometimes said, Jose, but you know, I don't mind what people said. You have the controls, it was well, let's interpret your results, okay? And now I was so biased I knew too much about the satellite

    and I had this thing there, you know, this, the tails, you know, it was obvious, you know what I mean? So I think that's the good advantage for this kind of systems, that there is no bias. The system, it doesn't matter if this is a phage, or, you know, it's a conjugation, you know, we always thought that the satellite should move by a phage because that's the name.

    We never thought, you know, but why not by conjugation? You know, it's just to put a very small sequence, yeah, so it was a good revenge, you know, in that way, you know, that the system, you know, gave me in that way because I knew too much. And I was, so biased. We were so biased when we talked about this thing, you know, and then, yeah, we realized. And another funny thing is that I have no idea about tails, okay, when, yeah, zero, I. It's something that I never, ever thought about that. The tails will be something interesting at all. That was a tail, it's form, you know, the funny things with the packaging, the capsid, you know, this is the big papers from the HK97 capsid, you know, the terminator, you know. And now suddenly it looks like the tail, it's a funny thing as well, you know, it's because, as you said,

    you can have now a very simple mechanism to move things between species, you know,

  82. Jessica Sacher

    wow, yeah. So was Google kind of excited that you guys showed the world a nice test case? Like, it's pretty neat how well it worked? Is it really that clear-cut? They must have been very excited.

  83. Tiago Costa

    Well, that's a question for them, obviously, but I think they were very pleased to have experimental evidence that the hypotheses the system was generating were sound and solid and verifiable in the lab. There were other studies that were initiated by hypotheses from the co scientist and are reported in the system's preprint. They still have to go through peer review and publication. In our case, we were much further along because we had already a full story, we knew the answer, and the system was able to recapitulate what we found.

  84. José Penadés

    We tried to take some stock options from them, but, you know, it's a joke.

  85. Jessica Sacher

    What was that?

  86. José Penadés

    Some stock options from Google.

  87. Jessica Sacher

    Oh yeah, I was gonna say they should be funding your lab.

  88. José Penadés

    they didn't agree with us. Okay, because I think that they were lucky, and we were lucky as well, you know, because I think this is the perfect system to test. Because, as I said, everything was there, okay, you can produce the capsid, you can produce the tails, whatever people have done experiments in the past mixing things, you know, tail mutants with capsid mutants, and then you have an infective particle. So all was there, just putting the dots with being biased. I think was lucky for them, and lucky for us, you know?

  89. Tiago Costa

    But it's definitely a very nice partnership.

    There was no funding involved, so they didn't give us a penny.

    so we need to be very clear with hands.

  90. José Penadés

    we are getting money from Google, we don't, we are trying, but so far no.

  91. Tiago Costa

    No, nothing.

    I think it was mutually beneficial for both sides.

  92. José Penadés

    I don't know if you checked the preprint, because we didn't put this in the preprint. Sorry, Tiago. But even the system provides key papers it used to arrive to that hypothesis.

    You think it would be nice, for example, to include that information as well for people, because they were mentioning names, as you know, and for each name, they highlight one paper that they thought was relevant for the hypothesis. And that was pretty cool as well, you know?

    For the hypothesis, they ranked, I think, a few papers.

    It was for the tail things, for the hypotheses they mentioned. There were fifteen papers that they said they found relevant to arrive at that hypothesis.

  93. Jessica Sacher

    And they were real papers, basically.

  94. José Penadés

    Yeah, real papers, and it made a lot of sense, you know.

  95. Joe Campbell

    Can you tell if the other AI tools that you used that weren't as effective? Did they see those papers and not realize the importance of them, or did they just somehow miss them? Or, I guess, in trying to understand what's different and why they're giving different answers, I guess I could broadly think of two categories. One is that somehow they didn't find the right papers, or they found them and they didn't realize, weren't able to understand the importance. So can you go back and figure out if those other search engines, even, or whatever learning tools or AI tools, even looked at those papers?

  96. Tiago Costa

    As part of the preprint, we had to benchmark the co-scientist against other systems, including Gemini and ChatGPT. The output from those systems is different. The format is not the same. However, the hypotheses and the answers are there.

    Well, the co-scientist outputs the references. It lists scientists who have worked in that field. So this is a very comprehensive report. Other systems do not give such a comprehensive output because they are built differently. I mean, we're not the AI people, and the co-scientist, at least for me, is still a black box. But the system is built differently. Funnily enough, we published the preprint of the experimental paper on bioRxiv, and then we challenged the different LLMs. Some hadn't seen the preprint.

    They could not even find our already published data, where the answer was written.

  97. José Penadés

    I think it was Google.

  98. Tiago Costa

    Was it?

  99. José Penadés

    No, I think the system found the preprint, but didn't provide the right answer.

  100. Joe Campbell

    yeah,

  101. Tiago Costa

    Or okay, so it's the same. He didn't provide the right answer, even though he found

  102. José Penadés

    But they found the preprint. I think was for one case, and in other systems, they also propose things that make sense. So, for example, some of the system propose that maybe this capsid forming can take tails that have the ability to inject DNA in different species, the same tail, okay, but, for example, in that case, that will work for both, for this capsid forming, or for any satellite, you know? So it's not something that is specific, because at the end, the tails are very similar. So some of the hypotheses were kind of okay, but none of the other systems provide the right thing. You know, is what different tails, and, you know, and even some of the systems. I can't remember which one was the last one we put in

    the paper. They made a very nice overview of the PICI cycle, you know? So it definitely had very good access to the bibliography. But I don't know to what extent, as I said before, that's a problem, because this is the problem we have. You have a very good, knowledge, then how you interpret things that, you know, it's something new. And that's the question that some people ask to us, you know, to what extent? These are just new hypotheses, or just linking, you know, or connecting the dots, and I think so far, it's more connecting the dots in a non-biased way. Way more than thinking something that is completely novel. Okay,

  103. Tiago Costa

    Okay,

  104. Jessica Sacher

    hm,

  105. Joe Campbell

    yeah, if I could ask just a couple more questions. The first is, are you in any way planning to rewrite your manuscript that you hadn't shown it based on what it did? So, are you? I guess, my guess would be it would most likely affect the discussion. But are you planning to rewrite it? Will you add to that before I ask the follow-up?

  106. José Penadés

    Right? For the experimental manuscript, nothing changed. What Google’s output addressed was already in the experimental manuscript. In the experimental manuscript, we also provide new concepts that the co-scientist never mentioned. We thought it would be nice to share our experience, so we are creating a new manuscript about our experience with the co-scientist.

  107. Joe Campbell

    right?

  108. José Penadés

    Explaining what's the input we provide, okay, and then evaluating, what's the output that we got, okay, and explaining, you know, for example, the tails, you know, this is a good thing, but this is things where you know. Explaining this kind of frustration or why we think a co scientist was able to arrive to that point, you know, even this non-biased, even did they provide the picture? And we also make kind of evaluation from the rest of the hypothesis. We preprinted that, manuscript as well, and we sent the manuscript to the journal as well, to the same journal that the experimental one.

    Because they might be interested in having both the experimental one and this

  109. Jessica Sacher

    love it.

  110. José Penadés

    co scientist related one.

  111. Joe Campbell

    I guess the other thing I was thinking about, and maybe this is a hard one, is if you'd had the output of this LLM system before you started the experiments, what would you have done differently, if anything? Would you have done different experiments?

  112. Tiago Costa

    Nothing changes. It's just

  113. José Penadés

    Quicker, much quicker,

  114. Jessica Sacher

    quicker.

  115. Tiago Costa

    the pace. Yes, so the way I see the system is, it's like a collaborator. It's something you can interact with. You obtain hypotheses in the end, but they are not the final truth. You have to go to the lab, run the experiments, interpret the data and draw conclusions. The hypotheses the system generates are not universal truth. The scientific method does not change at all.

  116. Joe Campbell

    right,

  117. Tiago Costa

    What it does, it puts you perhaps in the right path, right from the beginning. So if you think about it,

  118. Joe Campbell

    I understand that the hypothesis doesn't affect that. You need to do an experiment to address it, but I guess what I'm asking is, I mean, well, back when I used to do experiments, you know, often if you have a hypothesis, you design an experiment to test it. And were there things, so when you said it was quicker, is that because you would have designed the right experiments?

  119. Tiago Costa

    Maybe you would not have failed ninety percent of the experiments that you failed, but you would just fail fifty percent. And this would save you half of a year or one year of experimental work, and one year of funding, and one year of salaries, and one year of taxpayers' money.

  120. Joe Campbell

    So it would have changed which experiments you did and didn't do, right?

  121. Tiago Costa

    Exactly. It would change

  122. José Penadés

    with So I'll give you an example, you know? Coming back to the idea that we knew that we have some strands that we can induce. The island, it's been, you know, it's been induced, it's highly packaged, you know? And we tried to get transfer and we didn't get any transfer because there were just capsids, no tails there, you know? So can you imagine how many recipient strains did we try? Because we thought it was affecting the recipient strain?

    Okay, capsid without capsid, with these, you know, can you imagine how many? And we didn't get an answer for that, and then we moved to E. coli, the same thing, you know, and then start deleting these things, deleting, you know, because we never thought that these were tailless, okay, or we never thought that. Yes, if they have everything to package the DNA, just need a tail, maybe they can bind to different, you know? So it's like, at the end, we arrived to the right experiments, you know, but we failed a lot of the experiments because we didn't understand what was happening. You know,

  123. Joe Campbell

    right,

  124. José Penadés

    Yeah, if they give you some kind of path and the others. There are other hypotheses that you realize that might be irrelevant for the biology of the satellite, but they were not really important for the question, you know, because can apply to many other satellites, so there was no exclusive of this family. But the type of conjugation is the same. They provide the hypothesis that is very easy to test.

    Something that we highlight in this second preprint is that, at least in our example, all the five hypotheses are very easy to test.

  125. Jessica Sacher

    hm. yeah,

  126. José Penadés

    You know, so it was, and also, as Tiago said, we had the feeling that we were talking to an expert in the field. Even, you know, so it's what do you think about this? I think you should take this thing, okay, and then you, okay, that makes sense, you know? And definitely, you know, at the end, we were very happy because we arrived to this, to the conclusion, be, you know, the same one. But for the other examples, for example, the what the one that Tiago was saying, it's a mystery for seventy years and nobody thought about it. Okay, can you imagine how many experiments have people did in the past trying to understand the time zero for conjugation?

    This is a big question. I can't believe that these guys working on conjugation don't really know that. Okay, so, and the system provides you a hypothesis that looks okay, it's saving a lot of time.

  127. Joe Campbell

    Right, yeah. And I wonder if in your paper, about this method, you could sort of give concrete examples of how you would go from ninety percent of the experiments not working, to maybe only fifty percent of the experiments not working.

    And sort of say we would have done if we had known this before, we wouldn't have gone down this rabbit hole. And because, I mean, I guess, maybe I'm trying to think about if someone's out there trying to say, well, why am I going to do this? I mean, I think a lot of people would be attracted to know which experiments that you did, which ended up being dead ends.

    You think you might have just not have done if you had run this AI prior to starting the experimental work. Does that make any sense? Yeah,

  128. José Penadés

    Yeah, it makes sense,

  129. Tiago Costa

    yeah, yeah, It makes sense. You tend to forget the failed experiments and keep track of the good ones, right?

    I don't know. That's why, arguably, you don't publish negative results. Or maybe you should, but anyway, that's another question.

    But I think it's very difficult to say now which experiments should have been done, or should have not been done, or could have been done differently.

  130. Jessica Sacher

    It's like life. Sorry.

  131. Joe Campbell

    well, it's hard, but I mean, I guess it's a question of whether that highlights the importance of the

  132. Tiago Costa

    won't be an issue, but if you, if you think, if you think to the future, this won't be an issue. If it's an AI-driven hypothesis, you have never run those experiments before. Okay, so in this case, you, because we knew the answer, we have the record of failed and successful experiments, we could do that judgment. But in the future, if it's an AI-driven hypothesis of experiments, we will. I think that what will make a good or quicker discovery is how good the human will be to critically interpret the hypotheses and design an experimental setup. Although the AI system already gives you some experiments you could use to test that hypothesis, the human will still have a very important role in this process.

  133. José Penadés

    So I think I can answer your question, okay? So when you have this capsid forming, I don't know for how many years. I was almost ready to publish this capsid forming PICIs in two thousand and ten. Okay, and Richard Novick told me, Jose, you can't. Richard Novick is the person that discovered the PICIs in staph that he called SaPIs. This is the first member of the family, let's say he's the father of the PICI, he discovered the first element. Okay, and as I said, for many years, he discovered them, I think, in 1998, and for more than ten years, twelve years, people thought they were defective phages. So when we discovered this cf-PICI, probably in 2010. And he said to me, Jose, you can't publish this thing because then people will think that these are defective phages.

    They have everything, they have the capsid, they, whatever, so you can't, okay, so we wait. So it has been for many years that we knew they exist. They kind of a manual, you know, findings we. We knew that they were in many different species. We couldn't understand why a satellite want to use half of the genome to carry genes that they can hijack from the phage.

    We did a lot of experiments where we didn't see transfer. And the day talking to the student that we realized that maybe some satellites use a phage for induction and another phage for the tail. Everything makes sense, okay, and I think it's when Tiago was mentioning before, I said, Tiago, I think we have something pretty big, and since then, everything was very quick, very quick, because these experiments are very simple.

    If you have a capsid mutant, you can have the tails, you can mix, so these things even work in natural populations. So a normal phage prophage when it's induced, there is a lot of tails there. That if you just mix with the capsid forming, you have infective particles. Everything we have an experiment in the paper, we have a donor for the capsid forming, a donor for the prophage, a functional one and a recipient, and things move, you know, so, but that was the day that we realized that maybe this is something for induction and something to provide the tails.

    And maybe we realized that, you know, everything makes sense. So if you have something from Google that said, maybe this can use different tails, you thinking, you know, because the experiments are very easy to do, at least in our case, I don't know in other area. So I think it's right because of the bias. And I said, you know, it was a of all, because we didn't see the big picture, you know, but having something that is new is like the same thing.

    Conjugation. We never thought, I never thought about conjugation. Okay, Tiago never thought about conjugation. He's an expert on conjugation. There is a guy in Pasteur Institute, Eduardo Rocha, he's an expert on satellites and conjugation. He has made a lot of programs and studies about, OETs that are required for. And we sent an email saying, Eduardo, have you ever thought about this? and he said, Was a no, never.

    Because satellites use phages, plasmids use conjugation, that's, you know, this, you know, so we are so biased that things that are there, you know, we didn't realize. So I think that's the main, I think, the power of this system, at least for us. It has no bias. Whatever has been published, they will, you know, highlight that have you know, and then it's up to, as Tiago said, it's up to you to decide if that makes sense or not.

    And, you know, for the hypotheses they provide, we're very easy to test all of them.

  134. Jessica Sacher

    I wanna touch on that last point, I think the testability is really interesting. And I wonder if behind the scenes, the AI co-scientist that Google made, maybe they gave it a lot of training related to knowing what science is like. When, so it could understand, like, what would be a testable hypothesis versus not? Cause I don't think that's usually that's probably not, you know, what you're getting when you talk to regular ChatGPT or other LLMs, or even regular scientists.

    Like, I think that's like an advanced skill to think, you know, of all the things you could do, what are the things that are the most bang for your buck and the most testable? Least time in the lab? Like, you have to have a lot of background to know that, but even filtering through that lens to give you testable hypotheses.

    Like, it's pretty interesting that you said they're also testable, and I wonder if that's kind of like built into the system. And if so, that seems like a really nice feature. Is that your view of it?

  135. Tiago Costa

    The quick answer? I don't know. I think this is a question for the Google team. We do know the basic principles. The system generates different hypotheses. They are ranked, and then they compete and challenge each other. They are ranked with an Elo ranking, like chess players. We know it's a very complex system, because I've heard experts on AI talking about the co-scientist and they seem to be very surprised by the architecture of the system.

    And in particular, these novel hypotheses that have that Jose, which was mentioned about the oriTs and the that this cf-PICIs could also go, be mobilized by a conjugation. This what they call it indirect, a assumptions or hypotheses, and that they are not imprinted direct, or they cannot readily interpret by the data that is available.

    Okay, and so, whether there is an element of reasoning in the system that would come up with this, less evident hypothesis. Yeah, maybe, but this is, it's again, it's something for the Google people to dissect.

  136. José Penadés

    Sometimes it provides specific experiments, you know, for example, for the connector and adapter. They said you should check by cryo-EM how these things look for some viruses as well, the system suggest to use ribosomes. So it was quite funny. Even some of the hypotheses they found, even they were incorrect, they were quite funny because we included them in the paper as a negative. So, for example, one of the hypotheses was that just the capsid forming. There were proteins there that might have the ability to bind to some specific receptors without the tail.

    Okay, and we, you know, so it's like a different mechanism of entry. You say, well, this is a, and we include these controls because we have experiments with no tails to show that the tail was absolutely required. Okay, so,

  137. Jessica Sacher

    Okay, so you ruled it out.

  138. José Penadés

    Yeah, it was some kind of suggesting experiments, you know, in a very easy way, okay, but I again, I have the feeling here that the phage world is a very easy, it's a very good model for the system. Because it's just at least the work we do, the satellite is mutant complementation, this is some cryo-EM. Yeah, we don't really know in other areas, you know,

  139. Jessica Sacher

    I think we should probably let you go on with your evening and close out. But this has been so, so cool. My last question was, when can others use this? Are you still gonna be using it? Is it still in a testers-only phase? What's that like going forward?

  140. Tiago Costa

    The system is still under development, right? It is still not publicly available. This is again, that's the grounds of Google AI. But from our discussions with them, there's a process to develop the system before it gets released to the public.

    I think the strategy will be to, you know, to incorporate different areas of science into the system and understand how robust the algorithm is. And eventually, if it pans out as well as it has with the phage biology it will eventually be released to the public. I think what's written in Google's co-scientist preprint is that there is a stress-testing program, and the labs interested in testing the system can contact Google.

  141. Jessica Sacher

    Okay,

  142. Tiago Costa

    Yeah, we will definitely keep working with them, because we have a very close collaboration and partnership with them. And I believe it will be mutually beneficial for Google and for us as it has been so far. Hm,

  143. Jessica Sacher

    Yeah,

  144. José Penadés

    We're using it now with some very talented postdocs to challenge the system with the questions they really want to address, at different stages. So there are projects that are quite new, so to see how the system answers the questions and how the system's better established. This is a different approach. It's now asking for new hypotheses, you know, and we will give them time.

  145. Jessica Sacher

    The forward version.

  146. José Penadés

    Yeah, the forward version,

  147. Jessica Sacher

    I like your reverse version.

  148. José Penadés

    Yeah,

  149. Joe Campbell

    Yeah,

  150. Jessica Sacher

    Both make sense. It's like forward and reverse genetics, like the reverse AI co-scientist. It makes so much sense. Everybody else is sitting on data that hasn't put it out there.

  151. José Penadés

    we have found another thing that it's quite funny, you know, I don't know. So we work also with plasmids and the impact of transduction and plasmid mobility. Okay, so we have a paper now, again, nobody knows about the paper. So why do they have, pack or cost sequences on plasmids? Because if, as a plasmid, you really want to move, just put a pack or a cost, and then that's it, you will fly. Okay, so we have an answer for that. It's like, in reality, plasmids don't really want to move too much, okay, because then will, if there is a plasmid that moves too much,

    it will lose the variability or whatever. Okay, but this idea of the plasmids being less selfish than we thought still not, it's. It's a very small proportion in the literature. So far. For many years, we thought that the mobile are selfish. Okay, so when you ask the system why there is no cost or pack fit, sequences on plasmids, that body of literature is more important than the cooperativity or whatever.

    So most of the hypotheses that are pretty good are wrong because that's just based on the selfish, you know, they do these things, they the terminase will cleave it, the plasmid and the plasmid will not be functional. Or you know, what's more, on the thinking that was a selfish element, okay, it was something that block the biology of the plasmid, as planned. So it's a quite funny thing, you know, the thinking, it's really good, but yeah, because most of the literature is based on the selfish thing. You can see that from the hypothesis, okay, so the system also needs to learn probably what papers are more important. You know, the recent one versus, I don't know, so this is, we are working there, you know, in some, we try against the regards for the plasmid thing. The hypotheses were pretty cool, but they were wrong, And we are working now.

  152. Joe Campbell

    Did you try the other LLM systems with that one?

  153. José Penadés

    No. For this one, you can do that.

  154. Joe Campbell

    That would be interesting, I mean, I guess it could be scary, but it could be interesting.

  155. José Penadés

    no, no,

  156. Joe Campbell

    Depending on the question you ask, different LLMs work better.

  157. José Penadés

    That's a good point. And we are now having the chat and we're telling the system, Yeah, that's very cool, but you are ignoring that is a really selfish. You can have some kind of another type of relationship with the cells, so let me know what you think if you include that possibility as well. And we are waiting, you know, so this is, it's quite interesting, exercise, you know?

  158. Joe Campbell

    I guess you want to be selfish, but within reason, because if you're dependent on bacteria for survival, if you get too selfish, you might kill your own host. Like, I don't

  159. José Penadés

    Yeah, but for many years,

  160. Joe Campbell

    when I was a graduate student, I was working in a lab that studied this transposon Tn10, and it had this thing called multicopy inhibition. You would naively think that it just wants to keep jumping and making more and more, but eventually that's bad for the cell.

  161. José Penadés

    Right.

  162. Joe Campbell

    So it has a mechanism that makes it transpose less when they get too many in the cell. I was thinking about that when you were talking about it.

  163. José Penadés

    Yes.

  164. Joe Campbell

    Because again, that's another selfish DNA that you think is like, the best thing would be for it to jump whenever it can. But if you start jumping too much, you trash the E. coli that it's in. And then

  165. José Penadés

    Correct.

  166. Jessica Sacher

    Yeah.

  167. Joe Campbell

    All right.

  168. Jessica Sacher

    AI is gonna find Joe's thesis soon.

  169. Joe Campbell

    That was someone else's work.

  170. Jessica Sacher

    Oh, well, thank you all so much. And good luck with everything. And I can't wait to see your papers come out, and we'll be posting them. And yeah, can't wait to get this out. So keep it up.

  171. José Penadés

    Thanks a lot.

    Cheers.

  172. Tiago Costa

    Thank you for the invitation. Okay, bye-bye now.

  173. Jessica Sacher

    You're welcome.

  174. Tiago Costa

    Bye, bye.

  175. Jessica Sacher

    Bye.