Ep 140: Biology outside the box (with Oded Rechavi)

What’s the value of risk-taking in research? How is studying the mechanisms of transgenerational inheritance in C. elegans unorthodox and insightful? How can AI help improve aspects of biology, namely the peer review process?

In this episode, we talk with Oded Rechavi, professor in the Faculty of Life Sciences and the Sagol School of Neuroscience at Tel Aviv University. Our conversation with Oded was wide-ranging, starting off with creativity in science, then delving in his research on transgenerational inheritance in C. elegans. At the end of the show, we talked about q.e.d Science, an AI tool he and colleagues have developed to review and provide feedback on scientific research papers.

Cover art by Brianna Longo

Check out substack for bonus content!

  • Marty Martin  0:05  

    Howdy Cam! Question for you, discovered any new good bands lately? 


    Cameron Ghalambor  0:10  

    Yeah, there's a few new bands I'm listening to. I've been checking out this band called Oso  Oso, and I've been also listening a lot to Royel Otis.


    Marty Martin  0:18  

    Okay, but be honest, most days you'd probably prefer some good old fashioned dad rock, right? Creedence Clearwater Revival, Jimi Hendrix, Janice, Joplin, Santana. Don't usually go for those bands when you just want to relax?


    Cameron Ghalambor  0:30  

    Yeah, I listen to those bands. I'll throw in some jazz, some punk, some ska, some new wave into the mix. But what's your point? 


    Marty Martin  0:37  

    Well, I mentioned those bands in particular because they're some of the original acts of the first Woodstock.


    Cameron Ghalambor  0:41  

    Ah, okay, so before we go any further, let's make clear that we're talking about the original Summer of Love Woodstock in 1969, not all of the terrible attempts to take advantage of the Woodstock name, like that disastrous 1999 concert that was full of violence, arrests and deaths. The first one was historic. The rest were a travesty.


    Marty Martin  1:03  

    Okay, but are you sure you're not just against Kid Rock, maybe that he's not as talented as Arlo Guthrie in the Grateful Dead?


    Cameron Ghalambor  1:11  

    Not a fan of Kid Rock's music, but the issue with the subsequent attempts to recreate Woodstock had more to do with incredibly poor planning and organization. Even though there were a lot of amazing artists there, it was clearly a money grab. Anyway, now I know where you're going. You're setting up Woodstock of biology. It's that event that you went to this summer in Prague, right?


    Marty Martin  1:33  

    Exactly! This June, more than a hundred scientists got together to have an unconference, and they call that "Woodstock". At this unconference, speakers gave very short talks, just five minutes, but they also shared ideas and tried to push the fields forward in really different ways than we usually do at regular meetings.


    Cameron Ghalambor  1:49  

    So by "very", do you mean that you all gave talks out in a muddy field while tripping on mushrooms?


    Marty Martin  1:56  

    Uh no. And maybe it's best to just let one of the organizers the event and our guest today, Oded Rechavi, explain


    Oded Rechavi  2:03  

    There are a couple of ground rules was the first of all, everyone who wants to participate and signs up on time and they get to present. They get to present very short talks of just five minutes. And second thing is that every participant has a walk up song. They walk to the stage with music that gives a whole color and sort of makes it into a celebration, also shows you the personality of the different participants. And the third thing is that you don't know the old there's no schedule, so you don't know when you're going to present. And then the advantage of that is that in normal conferences, people often say, you know, maybe I'll skip the next session because I'll go get coffee, or I'm not that interested, or something. Here you don't know if the next speaker will be you, or, you know, the best talk in the conference, so it keeps people at the edge of their seat. And that also really works.


    Cameron Ghalambor  2:55  

    Wow, that sounds both interesting and bizarre. What was your walk up song?


    Marty Martin  3:01  

    Of course it was "Till I Collapse" by Eminem. My son, Alex, picked it for me. 


    Cameron Ghalambor  3:06  

    Hmm, I'm not sure about that choice. Yeah, I could see you being kind of more of a Raindrops Keep Falling on My Head, kind of guy. But we'll not go there for now.


    Marty Martin  3:18  

    I could tell our listeners about your obsession with Grace Jones. But I won't


    Cameron Ghalambor  3:22  

    You just did, but wait, I'm not obsessed with Grace Jones, am I? Well, okay, maybe a little bit.


    Marty Martin  3:28  

    Of course, you are. But anyway, focus the Woodstock of Biology event in Prague was actually Woodstock 2.0. Oded co-hosted this one with former guest of big biology, Itai Yanai of the Night Science Podcast, but also Petr Svoboda, Pavel Tomancak and many others, as Itai emphasizes on his Night Science show, science needs room for more creativity, and that's the kind of experience that they set up. Here's Oded again.


    Oded Rechavi  3:53  

    So the meeting was also organized together with Itai Yanai. He has a podcast on night science about the creativity in science so to include sort of a night element, or creativity aspect to the meeting, in between talks, we included special night science talks which were supposed to be these crazy things that people do. So, for example, Itai had to explain his science to a 16 year old who was on stage and was supposed to ignore him. Someone gave a talk with their hand puppet. There were all kinds of crazy things like that, and I think that also made it more interesting as well.


    Marty Martin  4:24  

    On today's show, we won't be talking to Oded about Woodstock, but we will post a link to the meeting's website on the Substack page, just in case you want to learn more. Moreover, keep your eyes and ears open for an announcement about Woodstock, 3.0 to happen in the next few years. Oded mentioned they're eyeing an old Romanian mine as the vineyard for that event.


    Cameron Ghalambor  4:25  

    A meeting in a mine. Wow, that could be a first. Anyway, we'll be talking with Oded today about the other very creative things that he's been up to. One is his research on epigenetic inheritance in C. elegans, which he characterizes as "radical".


    Marty Martin  5:00  

    And the second is a new artificial intelligence that colleagues and he are developing to serve as the ultimate reviewer two, an omniscient authority on the optimal study design, but also arbiter of data over interpretation.


    Cameron Ghalambor  5:13  

    They call it q.e.d. science, and we'll post a link to it as well on our Substack page.


    Marty Martin  5:19  

    Oded's day job is as a professor at Tel Aviv University, but many of you might know him from his years of social media work on Twitter or X and Bluesky. Many times every day, he has posted various pictures and short videos from movies and TV and other parts of pop culture, connecting them to the frustrations that we all endure with the business of science.


    Cameron Ghalambor  5:38  

    We loved our chat with Oded. His passion for thinking outside the box, or what Marty likes to call heterodox approaches to science, were really inspiring. Moreover, q.e.d. Science promises to be a game changer for many reasons.


    Marty Martin  5:50  

    Yeah, I've been using it extensively, as have my students and colleagues, and everyone is blown away by how much it can help improve a manuscript,


    Cameron Ghalambor  5:58  

    As you'll hear, it works well now, but it's just the beginning of what they have planned, so stay tuned.


    Marty Martin  6:04  

    And the last thing before we start the show, remember, we're in fundraising mode.


    Cameron Ghalambor  6:08  

    Our goal is 500 paid Substack subscribers, and since our last episode dropped, we made it to our first milestone, but we still have a ways to go. So thank you to everyone who signed up, but please sign up and become a subscriber.


    Marty Martin  6:21  

    Remember, for every new 20 paid subscribers, we're going to randomly select a subscriber to receive a big biology t-shirt or poster with cover art from any of our episodes.


    Cameron Ghalambor  6:31  

    Congratulations to Joe Morgan, who was the lucky winner of our last draw


    Marty Martin  6:35  

    For your own chance to win become a paid subscriber by going to big biology dot sub stack.com and signing up.


    Cameron Ghalambor  6:41  

    Once we get up to 500 subscribers, that winner will get a Big Biology sweatshirt and a large art print.


    Marty Martin  6:49  

    And, of course, those prices are all in addition to the benefits that paid subscribers already get, such as access to whole episodes, episode debriefs and extra audio from our guests about their lives and hobbies.


    Cameron Ghalambor  6:59  

    Please go to big biology dot substack dot com, and sign up today.


    Marty Martin  7:05  

    I'm Marty Martin.


    Cameron Ghalambor  7:05  

    And I'm Cameron Ghalambor.


    Marty Martin  7:07  

    And this is Big Biology.


    Marty Martin  7:21  

    Oded Rechavi, thanks so much for joining us on Big Biology today.


    Oded Rechavi  7:25  

    Totally My pleasure. Thank you for hosting me.


    Marty Martin  7:27  

    Let's jump into a dimension of sort of creativity and Cam found on your lab website that you sort of self identify as "radical". I don't remember exactly the title of the lab, but maybe you could see say more about that, relative to, I guess, what we could call mainstream,


    Oded Rechavi  7:41  

    Yeah, so I think, you know, radical also comes from "radix", the origin of the word, so goes deep into the roots. That's another something to it. But yes, I often am more interested in idea that startle. I would say, or you know, that bring you out of your comfort zone, or that are the opposite of what people think. First of all, it keeps me interested in the science, which is a challenge. See the chance. I mean, the science is interesting by also grinding and, you know, getting funding. So the talking with students is great, but then executing everything is not that great, always, not sometimes. And I think also the strategy, and this is something I often say to students and so on that. I mean, you never know where ideas are going to come from, good ideas, but a starting point could be, doesn't have to be, but could be trying to find something that everyone believes in and trying to ruin it or not challenge it. And the more people believe it, the more it is perceived as sort of a dogma, then if you succeed in even finding an exception, then it's always interesting. So I think it's a starting point for finding interesting projects. That's a good that's a good starting point for me. The line is that it has to be testable. And you try, you try. I mean, if you're a student of mine, try it, if you have the time, and be aware that you have a limited time in your PhD and so on, so don't do something that will never work. But if it has, but I think a sort of a good ratio is if more than 10% of your experiments work, then you're probably doing something not that interesting, I think. I think if it's one in a million, then it's too much.


    Cameron Ghalambor  9:23  

    Yeah, so, Oded, Marty and I talk a lot about the different ways people approach science and the different, you know, biases and the differences in how they approach questions. And, you know, I think saying up front that, you know, you want to do your laboratory for radical science that challenges the dogma. I mean, most people would not describe their research as like the laboratory for pursuing boring, accepted dogma, even if that's what they do. But it also seems a bit subjective. You know what is like radical to one person, and may not be very controversial for somebody else. Is this more like, sort of self motivation, as you were saying, like you get excited when you think that you know you're you're pushing the limits of, like, you know, what we kind of believe, and maybe the exceptions to what we believe, or is it more, kind of a way of like also selling and branding your research program to make it, you know, more high profile, I don't know, like, draw more attention to it?


    Oded Rechavi  10:35  

    So I think sometimes they people will, you know, they don't call it boring, but they would render the lab is something that doing something extremely technical for the sake of sounding very serious. So I think it gets close to that. But for me, it's, I think first of all the challenge, there are different challenges in every step of the career. For me, the challenge currently is staying interested and excited about the science. That's the main challenge. And I think that there are ways to quantify whether something is radical or crazy, so I think it's possible even if, very quantitatively, if you'd imagine all the claims that anyone has made in the world ever, and you sort of put them in a web of the claims that support each other or contradict each other and how science is structured in a certain way. And then there's a new claim and or a new subject or something you want to study. If it takes down the entire first floor and you need to rewire the entire net, then it is radical. And if it just changes the connections between the different arguments slightly, then it could be important, but, but it's less impactful or lower and also think that just very qualitatively, if you are if there's a long history of people saying one thing and you are questioning that, then it is more dangerous, and to me, also more interesting. And I think that also science often progresses from really just exposing holes in our thinking or contradictions, doesn't have to be, you know, full blown proper, but there is something to it.


    Cameron Ghalambor  12:10  

    Yeah, so isn't it Karl Popper's idea that, you know, who says that we have periods like, almost like stasis, where people kind of do the standard, you know, dogma, and then all of a sudden, there's a shift in thinking and move to a new, a new sort of equilibrium in our understanding of how things work. So I guess, you know, radical science, then maybe we could say is looking for those ideas that create these jumps in our in our thinking?


    Oded Rechavi  12:43  

    And I think philosophers, they argue whether really science progresses that way or not, in reality, but I think it's a strategy for choosing your subject of what you want to do, then trying to falsify something is a good start. Doesn't mean that always trying to do something very different is a good start. It doesn't mean that the history of science would really change in this way. Maybe there will be sort of paradigms of people working and they gradually change, are not replaced overnight. That's more about the dynamics of the scientific process. But as one lab choosing the subject, I think, trying to create science that's falsifiable and also that challenges what everyone thinks is a valid strategy.


    Marty Martin  13:29  

    How does what we're talking about relate to creativity? Because, I mean, we're sort of talking, you know, you mentioned that your pursuit of projects is about keeping the attention and staying interested. Is that sort of satisfied when something is creative or, you know, there's, there's a novelty to it? How does that relate?


    Oded Rechavi  13:49  

    So it's definitely about creativity that keeps you interested. And, you know, I think it's probably a matter of taste. And everyone is different. Some people, the stuff that gets them interested in going to the lab is that they want to cure a disease. They say, I have a meaning that I'm working towards the thing, even if it's the most boring thing in the world, and it's very incremental. Maybe we need a small increment to cure the disease. And this is what I need. This is what wakes me up in the morning. And that's great, but it's not really me. I mean, I just enjoy the creative process and doing things that are interesting to me. And I think that also science and advances in science are kind of random, including the development of drugs, and it doesn't necessarily come more from this concentrated effort on the path that is most likely to lead you to the drug. I think it's just informed by if, I mean, perhaps you should just randomly give grants to every scientist who wants to just try something. I mean, it's very possible that it would be more efficient than sort of a pharma company type of attitude that choosing one target, because there are indications saying that it could be useful. And we are just  putting all our chips on that. So you might as well enjoy the process. That's what I think.


    Marty Martin  15:06  

    Yeah, yeah. So do you think that we're doing? I mean, with that in mind, and I agree with you that a lot of discoveries are random. I even like how to award grants idea, maybe not all grants, but some grants, 


    Cameron Ghalambor  15:16  

    Except the ones for me.


    Marty Martin  15:18  

    Yeah except the ones to Cam. We'll just give all the money to you and problem solved. How are we doing by our students? Are we teaching them appropriately to be creative, and what kinds, if not, what kinds of things should we do more of?


    Oded Rechavi  15:28  

    It's a very difficult question, especially now with AI everyone thinking: "What should go into education? What should people learn, and what strategies should they have?" And very you know, no one knows, but very intuitively, I think it's reasonable to assume that the obvious things would be done but would be more automated, that AI would have an easier time doing the obvious steps. And maybe we should just go for the extremes, although the AI might be able to do the extreme as well, I don't know. But I think when it comes to our students, it's more about, the way I see it, it's about granting them freedom, and especially freedom to fail. And I think we're doing a horrible job at that. The architecture of academia is really bad for them. People are afraid of failure. You need a publication to finish your PhD, and all of these things and failures or mistakes cause you great humiliation instead of sort of understanding that this is the process. I mean, if you compare art and science or other creative pursuits, artists more than scientists, they understand that failure is part of the process, but for scientists, failure or mistakes are not and this is a great disfavor for their students.


    Marty Martin  16:53  

    So I think I agree with you. But then, in putting into practice that agreement, what can we do differently about how we train? Let's take our graduate, our PhD students. How could we enable them, or sort of not facilitate failure, necessarily, but just, you know, accept that it's part of the process and bake it into the pursuit of the degree?


    Oded Rechavi  17:11  

    So for example, people say, you know, you need to have, you have this crazy idea of yours, fine, but you also, as a, you know, PI I'm telling you, as a professor, I'm telling you, you also you must have a safe project. By safe, we mean boring, right? A safe, boring project that you so you can, you can finish your PhD, right? Why? I mean it's, it's not necessarily needed. Imagine we didn't require publications to finish, or we would also, you'd say, even better, of course, just publish negative results or complete failure, which is something we need to do. That would be great. And I think we'd have more exciting discoveries. In some fields, there's this thing of registered reports. I don't know if you know this thing, but so, but it's not popular in biology. In psychology, I think it's more popular. What you do is that you submit your plan for the experiment, and then the plan gets reviewed by the journal, and if it is accepted, to my understanding, then the paper is accepted. But you have to do the experiments. If you don't do the experiment, if you don't publish it after a while, then it's sort of a retraction, I think. And that's sort of meant to increase publication of negative results. The problem is that science, often, in biology, at least, doesn't always work like that. You wanted to do one thing, but then the experiments led you in a totally different direction. So maybe this is a better for a simple experiments in psychology, when I'm going to do one manipulation, compare this group to that group, but in biology, molecular biology, it's really not how science is done, but some ideas, some as some ideas for publishing negative results, for encouraging failures, I think that would be helpful.


    Cameron Ghalambor  19:01  

    Let's now transition to talking about your research, your radical research on C. elegans. And specifically, I think you know, we want to talk about the inheritance, the transgenerational inheritance, of these small non-coding RNAs. Maybe tell us a little bit about what are small non-coding RNAs? I know they come in many different forms, but you had a Nature Reviews paper where you kind of talked about some of the diversity. And so maybe for our audience, can you just give a broad, brief overview of what these small no- coding RNAs are?


    Oded Rechavi  19:39  

    Absolutely so small, non-coding RNAs are short sequences of RNA, typically between 20 and 30 nucleotides long, but there's some variation. And what they do is regulate gene activity. They do it in many different ways so they can regulate transcription. They can regulate translation. In most cases, almost all cases, they work by inhibiting gene expression, not really by leading to the expression of genes, mostly by inhibiting it. Originally, most of the studies focused on the capacity to slice or destroy the mRNA, but they also prevent the transcription of mRNA. And they do it in many different ways. There are all kinds of mechanisms when it comes to the action, the inhibiting action in the cytoplasm. In general, I would say that people refer to the action of small RNAs as RNAi or RNA interference, or gene silencing. When it comes to the capacity to inhibit they do it in different ways. They can lead to the slicing of the mRNA. They can lead to this destabilization of the mRNA. And in the nucleus, they can also change modifications of the chromatin, induce the deposition of repressive chromatin marks, or also affect the polymerase, stalling it, preventing it from elongating. They do all of that, not on their own. They also, they always do it owing to proteins that carry them. And there are many proteins that affect small RNA biology, the partners, the close partners, are called algonauts, and they are the ones that carry the small RNAs, the short RNA, and can execute their functions. They can either slice their targets, or they can guide the deposition of chromatin marks and so on. And there are endogenous, smaller RNAs and they are exogenous. Small RNAs that can be can come from the outside.


    Marty Martin  21:40  

    Okay, so, I mean, give us a sort of example of the context in which these things are produced. I think every biologist knows that gene expression is sort of how you produce phenotypic variation. I'm leaving out loads of details, of course. But is it in the typical operations of a cell that these small RNAs are produced? And if so, why? I mean, couldn't you just turn up or down gene expression? Why bother to have this other secondary mechanism to fix your mistakes? Or where did this process come from?


    Oded Rechavi  22:11  

    Well, this was discovered. It's a convoluted history of discovery, but originally people saw that when you apply double strand RNA from the outside, outside, you bring it into a worm, and you can just do it by feeding or by soaking the worms in double strand RNA, or by injecting double strand RNA. These double strand RNA will be chopped into small RNAs that then will execute this silencing effect. Okay, but later, they found that also they exist naturally. There are all kinds of endogenous small RNAs that are being transcribed. Most famous ones are probably microRNAs. This year, Victor Ambros got the Nobel Prize for the discovery of microRNAs. And these are smaller RNAs that are transcribed, in most cases, with exceptions, as genes. So they have promoters. They are being transcribed as a longer sequence, which is then chopped up and processed into these microRNAs that have specific targets, sometimes many, sometimes not so many. But there are many additional small RNAs that come from all kinds of places. There are small RNAs that process from double strand RNA. There are smaller RNAs that are processed from other genes. Some micro RNAs, as an example, are derived from introns. There are small RNAs that arise from transposable elements. So they come from all over the place, and it's an additional layer of regulation. And as you said, Marty, there are many ways to regulate gene activity, and I guess the cell needs that too, because the secret is really coordinating the timing and location of gene expression, and you need, apparently, many, many layers of regulation. I think when it comes to small RNAs, something which is typical is that often they evolve. And perhaps the whole system has evolved as a sort of a way to repress or fight mobile elements such as transposons or RNA viruses. And that evolved into something that the cell uses for many additional things.


    Marty Martin  24:08  

    Interesting. Okay, I've become really excited. And I know Cam is too about transposons so we have to be very careful not to move into that space.


    Cameron Ghalambor  24:15  

     Yeah. So let's talk about one of your I guess. Now, a little bit of an older study back in that was published in Cell back in 2014 where you expose C. elegans to starvation, food restriction, and C. elegans has this very interesting aspect of their biology where they can, kind of like shut down if they're exposed to the to a lack of food and then resurrect themselves later on. And so you did this type of experiment, but then you found that the recruitment of these small, non-coding RNAs, and the effect of these small non-coding RNAs was actually passed on not only one generation, but I think three generations from the parents. So can you tell us a little bit more about this study? And I think this is one of the first demonstrations of trans generational inheritance, of this type of environmental experience by the parents.


    Oded Rechavi  25:21  

    Sure, so I'll start by maybe a broader view a little bit, which is that you know, you need to understand why is this radical? And some of the, some of the disappointment for me is that it's not that radical anymore.


    Marty Martin  25:36  

    You can take credit for that, right? Yeah, you've de-radicalized.


    Oded Rechavi  25:39  

    I'm not the only one. I'm not the only one, but yes, sort of. The idea of transgenerational epigenetic inheritance. And if you want to be more provocative, a bigger thing is also inheritance of acquired traits, which goes back to Lamarck and maybe the Greeks. It's very old. And became, sort of the one thing that you can't say to biologists, that can't be true. And there's a very long and controversial history. When RNAi was discovered in C, elegans, and this is the work of Fire and Mello, they got the Nobel Prize in 2006 for the mechanism of how double strand RNA leads to silencing, to the production of smaller RNA and silencing. They found already in the first paper, the Nature paper from 1998 that when you accidentally inject the worms, or not accidentally, I mean, the legend says that it was an accident, but you know, in the paper, it's not accidental. But when you inject the somatic cells of the worm, you get silencing not only in the tissue that you injected, but you get silencing throughout the animal's body and also in the next generation. And we now know that there are mechanisms for shuttling RNAi effects between different tissues, including from the soma to the germline, which goes against one of the basic rules of dogmas of biology. But this was conducted using double standard. It was in artificial double strand RNA, double strand RNA that was introduced to the worms from the outside. 


    Oded Rechavi  27:23  

    In 2011 before even the paper that you mentioned. We showed, that was when I was in my postdoc with Oliver Robert in Colombia, that when the worms are challenged with viral sequences, small RNAs are produced against these viruses. The worms don't have dedicated immune cells pieces all of that. They use RNAi to ward off viruses and do it, and they do it very, very successfully. And we show that this also leads to transgenerational inheritance of immunity against these viruses. Later we showed, and this is the paper that you mentioned, that also when you starve the worms, it creates a trans generational effect. The descendants are different as a result. And this is because of small RNA inheritance. We know that because we now know a lot about the mechanisms of RNAi inheritance, including we know genes that are required only for heritable, transgenerational RNAi, but not for RNAi within the same generation.


    Oded Rechavi  28:15  

    So for example, the most famous gene out of this category, there are many now is called HRD1. This stands for heritable RNA deficient one. And this is an Argonaut, it's a protein that carries smaller RNAs, which is required for passing the silencing effect to the next generations, but in the parents, it still works. There are others like that. So the transgenerational starvation responses, which we studied, but also many other labs have studied different paradigms of starvation is dependent on RNA inheritance. 


    Oded Rechavi  28:54  

    And the reason we went originally for starvation is that the most well-known study in humans about sort of heritable effects is this Dutch famine or hunger winter study, where people studied descendants of people that of women that were starved during the Second World War in the Netherlands, and the children were affected. I would say that this is a big epidemiological study, there are many papers published about it, but it's not necessarily really epigenetic inheritance, because, in that case, the mother is starved, but the child is already in the mother, so it's in utero, and the child already has germ cells. So you're really affecting two generations down the road directly. Only when you go further to the F3 or the descendants that are not affected by the starvation themselves, you can. You need to evoke some other mechanism. And this wasn't done yet in that study, in this study. 


    Oded Rechavi  29:52  

    This is what we did in C elegans, in that study, where the starvation, but also in other studies, we studied the effect of high temperature. We show that the small RNA mediated effects, heritable effects, can last for many, many generations. So they are really affecting descendants that were not exposed to the original trigger. And there the, I think the reason that it's not that controversial anymore is, first of all, many have seen already that RNAi works transgenerationally in C elegans. There's really tons of labs showing that. But, in addition, we understand the mechanism. So here we understand there's a big claim. It goes against what people thought for 200 years, but there's a mechanism which is now studied and understood. In C elegans, one of the secrets to transgenerational inheritance is that they have these enzymes that are called RNA-dependent RNA polymerases, which amplify small RNAs in every generation. So what happens is that the small RNAs, they target an mRNA. And then it's a complicated process I can I can say a little bit something about it. What we now understand is the mRNA is being chopped, an enzyme comes and add this sequence, untemplated sequence of U and Gs, sort of a tail, which serve as this tail. The longer it is, the stronger the silencing and and it serves as sort of a landing pod for the RNA-dependent RNA polymerase to then come and land on this mRNA and run across it and use it as a template for generating more small RNAs, and then these smaller RNAs can come again and chop more mRNAs and so on. So it's sort of as a feed forward loop that maintains the silencing so the smaller RNAs don't run out. I mean, otherwise you think about everyone produces hundreds of babies, it would dilute extremely fast, but it can last for a long time. 


    Marty Martin  31:42  

    Yeah, yeah.


    Cameron Ghalambor  31:43  

    Well, there was something I think that you said that I think is really, really important here. And I just want to make sure that I understood correctly. So did I catch this correctly that there are two classes of these small coding RNAs, those that act sort of within the generation, and then those that are specifically for this kind of transgenerational effect is that, did I catch that correctly?


    Oded Rechavi  32:08  

    I wouldn't say it's necessarily the type of small RNAs, but there are effects that are that affect only the parents and there are effects that are heritable, and whether this silencing would transmit to the next generation or not depends on many factors, many of which we don't understand, some we probably understand. For example, transgenerational science in most almost all the transgenerational silencing that we know affects genes that are expressed in the germline. And this makes sense if you think about the mechanism, because you need the template of the gene in the germline so that the smaller RNAs can amplify. There are some exceptions which are very interesting, and we are studying them. In C. elegans, it's clear that silencing is moving from the soma to the germ line, but in the next generation doesn't move out of the germline to the soma, except in very rare cases that no one understands. Okay, so for example, if you have a worm that expresses GFP, green fluorescent protein all over the body, some ubiquitous promoter that drives expression of GFP in every cell, and you treat the parents with double standard RNA to induce silencing of GFP, in the parent, it will be silenced all over but in the next generation it will only be silenced in the germline, but it will be expressed again in the in the soma. Okay, why? It's an open question. We are studying it. I think we have some clues. But, you know, we're not done with that yet.


    Marty Martin  33:32  

    Yeah. So this is all remarkable. The two things that really stand out to me, because I'm gonna be the curmudgeon. So to what extent is this a C elegans phenomenon? I mean, this is something that seems well nailed down, but is this just a quirk of this, this worm?


    Oded Rechavi  33:48  

    So I think there are a few, a few big questions. This is certainly one of them. I think that there are some people who brings up arguments, legitimate argument for why this would be a C. elegans thing. For example, they say it makes sense to transmit parental responses across generations to the next generations when you live a very short life, when your generation time is short, because then the likelihood that the parents and their children share the same environment is high. So it's worth preparing the offspring what you experience, and also in animals that don't go very far, like, you know, sea lions or maybe plants. So kids, you know, end up living in a very far away country, in a totally different world run by robots. Maybe it's not worth preparing them for what we experience. I think this is it's an argument. It's fine, but I think that it all depends on the scale of how you look at it. Some things that we encounter, like, for example, pandemics, they retain over and over, beyond our lifespan. So, so okay, that's this argument. 


    Oded Rechavi  34:54  

    The second argument is that is mostly about them, the special machinery that C. elegans has. These RNA-dependent RNA polymerases that allow them to amplify these smaller needs, which we showed in 2011 are required for RNA inheritance. So they're not there in mice and in humans and in flies. They do exist in other organisms, for sure, in plants, some fungi, scorpions, bats, without it. So that's interesting, bot many people know, but true. People say it can't happen in humans because we don't have RNA-dependent RNA polymerase. I think this will be proven probably wrong. I mean, we now, we now have in the lab an example of an animal not published yet, but hopefully soon we will, you know, pre-print it probably, I thought it was we were supposed to public to pre-print it five years ago. It's still going on. It doesn't have an RNA-dependent RNA polymerase, nevertheless, a very strong transgenerational silencing, and I think that the reason is there are many feed forward loops that can maintain it. For example, C elegans don't have DNA methylation. In plants, small RNAs also induce DNA methylation, and this is part of a loop that then leads to the more generation of smaller RNAs. So there are all kinds of feed forward processes that can perhaps perpetuate, it doesn't have to be RNA-dependent RNA polymerase, but these are the arguments. 


    Oded Rechavi  36:15  

    And I think that another, if you want to, you know, people you know, say that there must be other there must be aliens somewhere, because the galaxy is very big. So I think here again, the chances are that we landed one of the five main model organisms of the world of biology has it, and no other animal has it, seems unlikely. And also, I think that if you just think about the effect about the animal world, a big percentage of the animals on this planet are worms. So that tons of worms. Some estimates say that four out of five individuals, if you just count individual, is a worm, so the effect on animal life in general, could be very big and in plants, we know it also happens.


    Marty Martin  36:59  

    Yeah, yeah. So another question that I mean, I think I'm with you about the generality. It seems reasonable to expect that other things are there when you don't find the mechanism that's seen in C. elegans in other taxa, that doesn't necessarily mean the phenomenon doesn't happen. This was invoked forever in immunology, where, you know, things without backbones couldn't possibly be sophisticated. Immunologically, they don't have memory. They can't have memory. They have no B and T cells. Well, that doesn't mean they have don't have memory. It means they don't have B and T cells, right? And so that's since been vindicated for a long time. Most invertebrates have some kind of memory of past infection, as do bacteria and on and on and on. But over how many generations, what's the limit that you've pushed? How many generations have you seen this inheritance endure? Because at some point it would seem to me that it starts to become adaptive to forget whatever your great, great, great great grandparents did. And so have you made predictions? And do you sort of see any intelligible patterns about the rate at which they stop? The germline stops paying attention to what it's being told to do?


    Oded Rechavi  38:00  

    Absolutely, absolutely. We studied it a lot. We saw a couple of very interesting things. First of all, most of the transgenerational effects to in response to just exposure to RNAi, to double strand RNA, which is the easiest thing you can do, because then you know that you're silencing one gene. You can just measure the silencing of this gene. If you do something complicated, like starving the worms, or, you know, changing the temperature, then many things happen, and it's harder to keep track. But the most, the most quantifiable thing you can do is trigger, the silencing of a particular gene, like a reporter gene, a GFP or something, and then see how long does it last? In most of these experiments, it lasts for approximately three to five generations at the population level. At the population level, this is not explained by dilution. After five generations, it's already in the billions. It's totally homeopathy. And there are mechanisms probably that control the timing, and we studied it a little bit. I would also say, before I explain a little bit about that, I would say that there are also cases where the effect lasts for a very long time. And we and also others documented cases where something goes for 500 generations, or, you know, or a few years. 


    Marty Martin  39:13  

    Wow


    Oded Rechavi  39:14  

    That's an advantage of, you know, studying C. elegans. A few hundred generations is, you know, within a PhD. Try to do it in a mouse or something. And we also did experiments where we just try to look at patterns of segregation of transformational responses in populations. And so we started with a single mother. This was the work of Leah Houri-Zeevi. She published two papers on the timing of transverse responses in the lab in a Cell 2016, Cell 2020 and so she quantified, for how long does the effect last, and how does this segregate? And we were hoping, you know, to find something like Mendel's laws. We didn't find Mendel's laws. Okay, it's not that elegant, but we did find some pattern. Ones that explain for how long does it last, and or allowed you to predict for how long would it last and how it would segregate. And I think it's a little bit too complicated to explain now.


    Oded Rechavi  40:10  

    I'll just give you one example, which is very cool. I told you that when you start an RNAI response, it lasts, typically at the population level, for three to five generations. There would be individuals, however, that would go and carry silencing for much longer and period. And if you select them, for example, you can maintain for a long time. Now, what we found first, this was Leah's discovery in 2016 that if you take the ones in your science, one gene, let's say, GFP, and then you'll see these dynamics. Three to five generations, most of the worms lose the silencing. And part of it is also that you are limited as an experimentalist. I mean, you can, you can only track so many worms. But what she found is that that's very cool, is that if you silence GFP, and then in the next generation, you challenge the worms with double strand RNA. Now you silence a different, unrelated gene, let's say a gene that M herry, that leads to expression of red fluorophore. Then it would extend the silencing of the GFP of the original silencing so. And you can keep challenging them with more double standards against other genes, and they would silence for longer. And this is partly because the whole thing is regulated at the systems level. There's a competition with the preference of the machinery to inherit, to amplify and transmit inheritance of smaller needs that came from the outside to the production of small RNAs from the inside, these endogenous small RNAs. And here you tilt the balance towards the this exogenous system, which is why it is extended. So here's one sort of thing. 


    Oded Rechavi  41:52  

    We also see all kinds of patterns that we don't understand. The first to show it really was Ben Lerner in a paper in the science in 2017. He studied a response to temperature. He showed that when you put the worms in high temperatures, it changes the expression of reporter genes in the next generation. And if you put them in high temperatures for one generation, then the response would last approximately five generations. But if you put them in high temperatures for five generations, now that the change in expression would last after you remove them from the high temperature for 14 generations. So something accumulates, 


    Marty Martin  42:25  

    Wow. 


    Oded Rechavi  42:26  

    Okay, there's many things there that we don't understand, but there is a mechanism for sort of counting, or something affects the duration. Another interesting thing that was, thing that we did, we published a paper in eLife, I think is a 2021, I'm not mistaken, where we show that if you, if you silence a gene, typically it would last for three to five generations. But now, if you stress the F1s in all kinds of ways, you either starve them or put them in high temperature, or put them in high salt, all kinds of stresses, then it will stop the inheritance immediately, sort of, and this is sort of a forgetting mechanism, perhaps. And if you want to rationalize it, and it's, it's hard, because it'll be a bit like evolutionary psychology, you know, something like that. But if you try, if, if you try to do that, then you, then you say, Okay, now if they started this response, but now the environment changed dramatically. They say, we need to find a different direction, so maybe they reset it. And it could be adaptive and it could also be maladaptive, which is something always important to stress. Many of these transgenerational response are maladaptive. So for example, the most studied transgenerational response to RNA in C. elegans is this accumulated sterility, which is probably not adaptive, but so 


    Marty Martin  43:41  

    Not so much.


    Oded Rechavi  43:44  

    Although, you know, probably not so much. But this is the annoying thing about biology. You could also imagine that, for example, worms that have less progeny, perhaps would live longer. Maybe it would be good in some scenario, many things could happen so it's complicated.


    Marty Martin  44:01  

     Yeah, that's the fun of biology.


    Cameron Ghalambor  44:03  

    Yeah, okay, so I have, I have so many questions, let me, let me start with the sort of a two part. I think what you just mentioned about the adaptive versus maladaptive sort of consequences of this is, is very, very important, because my impression is that most people who are interested in these kinds of transgenerational effects assume that the these transgenerational effects are always adaptive, because it's some way that the parent is able to pass on information to the next generation that, you know, may not be within the structure of the DNA. So my first question is just to clarify, so we have this very strong signature when we look at the patterns of inheritance with these small, non-coding RNAs, but when you look at the phenotypes, are the phenotypes also persisting in the same way, like in parallel? And the reason I ask that is because we know that gene regulatory networks are multi-dimensional, and organisms have the capacity to sort of reroute and get the same phenotype using sort of different mechanisms. So like maybe your parent experienced starvation, and you inherited these small, non-coding RNAs, but all of a sudden you're in a high food environment. Do you have the capacity to to get around these, these effects to, you know, change your phenotype in a in an adaptive way?


    Oded Rechavi  45:36  

    Right so I think that the problem is many people reported phenotypic effect that seems to play the same dynamics like, say, three to five generation, a number that comes up. However, it's tricky, because unlike this simple assay, where you just silence a gene, like with a fluorescent molecule or something, and quantify it, and you can do it very well. When it comes to the effect, the complicated effect of stress, then it's hard, like measuring lifespan, it's difficult, and it's also at the population level, or measuring the capacity to withstand heat, shock, all of these things like that it's not so easy to do. So we don't really know, and it's very possible that some since some cases, it won't be exactly the same, that perhaps small RNAs will still be inherited, but you won't see the phenotype, or perhaps you'll see a phenotype, but at this point, it will be maintained by a different mechanism, like chromatin modifications or something like that. So it doesn't have to be a one to one. And the truth is that we don't really know yet to what degree it overlaps. And I will just add, regarding your remark about the adaptability, or whether it's always good, it's not always good, that's that is true, and I think that is also understanding this is also a way to get out of this teleological trap, as if the worms knows what's good for it in the next generation, no. Some responses carry on because of these mechanisms, and then natural selection, besides, it's more about diversifying your responses or your phenotypes in the next generation.


    Cameron Ghalambor  47:08  

    Yeah, because I think what, what is really interesting for me is that, so you know, organisms have this system of DNA to pass information on from one generation to the next, and then within generations, we also know that organisms are plastic, and they have mechanisms to modify gene expression and change their traits in response to different kinds of environmental variation. So then to add this other layer of transgenerational inheritance, and for it to be stable, seems like you know they're, they're, they're, you run the risk that from an if it was truly adaptive, that the parents could be actually creating a situation where the offspring are are maladapted to their to the environments that they experience. And I know there's a paper by Tobias Uller who looked at how frequently the environment of the parent predicts the environment of the offspring, and actually, in most cases it did. It wasn't a very, very good job. Maybe C. elegans that, like you were saying, have this very localized movement, maybe more like plants, maybe an exception to this. But I think for those of us who study, you know, birds and insects and things that are moving around a lot, it's hard to kind of understand, like, why you would get locked into a certain pattern of inheritance, aside from what's already in the DNA? Does that make sense?


    Oded Rechavi  48:42  

    Yeah, and I think, I think it's, it's one of these situations where it's hard to make predictions. I mean, yes, you can say something like that, but you know what about, for example, pathogens that you meet over and over again? And we know that this is very influential in evolution. I think that making, making predictions about what would be efficient or what would be is just really hard. It's just, it's like people thought that it's impossible that the nervous system would be, that neurons would communicate through chemicals, because it's too slow. But then it ends up they do. So I think it's just extremely difficult to foresee all the possible scenarios.


    Cameron Ghalambor  49:22  

    Yeah, yeah. No, that, I agree,


    Marty Martin  49:25  

    Yeah. I mean, if you take it, I think. Cam, if you take your logic to the extreme, doesn't that sort of argue, argue against the utility of DNA too?


    Cameron Ghalambor  49:33  

    Well, you know, I guess there's, you know, DNA, again, you know, I thought another really important point that you brought up is that, you know, for something like C. elegans that has a very fast generation time, DNA should be very efficient, because you have a you have the ability to to evolve very fast, as long as you know there's genetic variation in the population. So, you know, I think you have. Something that's very stable, that works over, like, across generations and over longer time periods, and then you have, like, within generation, the capacity to then fine tune, and so that that's, that's kind of what I the way I kind of see things, and I guess, why I struggle,


    Oded Rechavi  50:18  

    Yeah, I think, experimentally, we don't know. For example, we don't know. I mean, C elegans are hermaphrodites, self-fertilizing hermaphrodites. We don't know whether these happens more in organisms like that. You can imagine, perhaps, why yes or no. And many of the conclusions come from absence of evidence. As Marty said, you know, you don't find it in mice, we say it's impossible it would happen in mammals. But even in C. elegans, the lab strain that everyone works with comes from England. In that strain, there's RNAi inheritance, it's obvious. But in many wild strains that come from other places, many strains come from Hawaii. There's RNAi, but it's not inherited. Or in C. briggsae, a related nematode, it doesn't have the capacity to take up our double strand RNA from the outside, C elegans readily does it. But you can easily fix it. I mean, you've just expressed one transport and from C. elegans in C. briggsae, now it does. Now it works. So I think RNAi, smaller and small RNA biology is very fastly evolving. It's in arms with transposable elements. It's any changing and very fast.


    Oded Rechavi  51:28  

    And for so one striking example that's related to evolution of that is that instead of this in yeast, they don't have RNAi, but you can make it work if you just express two genes, Argonaut and Dyson. And that works.  But they have a paper where it's a one figure paper, where they've shown that the loss of RNAi  in Cerevisia is the recent loss in evolution. If you look at the tree of yeast, most of them does do have RNAi. Cerevisia lost it, but they did it to coexist with a virus that's called killer. So they want to live with this virus, because it grants them with some advantages, kills their competitors or something. So they neutralize RNAi, so that the virus can propagate within them, and then they have this advantage. Otherwise, the RNA would neutralize the virus. But they showed in that paper that this is sort of a cheap trick that is like thinking short term. In the long term you want to have RNAi. And maybe there will be some local consideration, or short term consideration will make it worthwhile not to have it, but in the long term you want to have it. So I think it's really complicated. It depends about the ecology, and, and, and this is the stuff that we don't know at all. So we don't know whether it's, how does it work in a sort of ecological and evolutionary frameworks and so on.


    Marty Martin  52:50  

    Ah, well, typical biology so incredibly diverse, almost intractable, wonderful.


    Marty Martin  53:01  

    Well, let's, let's maybe turn the page Oded, for sake of time, because we want to talk about this very different, in the sense that this is not radical science. This is a radical approach to thinking about the sort of science as a job, in particular, this q.e.d science initiative that you've started. And I want to spend, you know, the rest of our conversation on that, but I think we'd be kind of crazy not to talk about this third hat that you've worn in terms of talking about the publication process and the business of science on social media. So, you know, Cam and I first became introduced to you, you know, about these kinds of things, and then discovered all of the amazing C. elegans work. But how do you find the time to do this? And why did you decide to make sort of commentary on how silly publishing, in general, in science is? I mean, why spend so much time on that?


    Oded Rechavi  53:53  

    Yeah, so, so I think it's not strategic. I mean, you have to think, I mean, I started so, so I started on Twitter when it was still Twitter in 2016 and back then, I had a publication published, and I wanted to promote it. So I did it, got maybe, you know, 50 views or something, it was really not successful, but, and I thought it's not for me, you know. But still, I kept at it for some reason, because, I mean, not for some reason, because this is what they do. They do. They know how to addict you to their apps. And I started making jokes about the frustrations of science and science publishing in particular, not only and that sort of picked up, and people liked it, and it became a thing of its own. And then it went out of control. I didn't really even want to do it. I just liked, you know, I think we all that when we when we meet each other, and we meet other scientists, we make joke about the process, because you tell it, if you tell about it to a normal person from the outside, I mean, they are shocked "what's going on?". And part of is it's not necessarily that the publishers are evil or something. I don't think that's the reason. It's also it's it's archaic. These are relics of how the system was a long time ago and there are many reasons, but it's not only publishing. Academia in general is weird. The way we choose our colleagues, the way we do hiring everything. From the student level to the professor level. So the jokes that I made about it, many of them are very stupid. They appeal to the people from in the scientific community, because normally no one knows about it. And I've been making these jokes for many years, for nine years, my calculations are correct, okay, and only now I built q.e.d, which is about that idea to change this problem, to challenge these problems. So you'd have to think I'm a genius to think about it nine years ago, I didn't, I really didn't know, of course, and of course, no one thought about AI, at least not me. Probably some people thought about AI, but I certainly didn't think about AI, and definitely didn't think about AI as a solution for this nine years ago. Okay. 


    Oded Rechavi  56:07  

    So the evolution of QED, which we will discuss, which is an AI for trying to improve the situation. It was natural for me, because these things interest me. I like talking about it. I like thinking about the meta of the science world. Because, I think, as we discussed, we discussed in the beginning of, how do you improve creativity in science? Part of it is about the architecture and how science is done. Not only about the topics, it's all about the process. And one thing, in my opinion, that really reduces our joy as scientists, because I think, I think we have the best job in the world, don't get me wrong. It's a hack, a hack in the universe. Okay, this is also important for the students in the beginning of their career. I think it's a hack in the universe now. And to this morning, I was with a friend of mine, a professor from another university. We just went to the beach to talk, you know. We swam at the beach. It was great. So this is something we can do, as scientists and most people in normal jobs can do. We have unparalleled freedom. But I think that one of the things that reduces from the creativity and the spontaneity of it and the fun of it, and also biases science, is how we publish, what is a paper, and how we write papers, and how we evaluate papers, and how we read papers. So this is why we're doing q.e.d.


    Cameron Ghalambor  57:25  

    Okay, so on on the website. I think the quote here is that the idea, in a nutshell, is to build a strong AI reviewer that provides authors with deep, insightful feedback, and that on paper, as you were saying, academia is a dream for curious, independent people, but in reality, it's publish or perish cycle with peer review at the heart, which actually reduces creativity. So AI is something that people are having a love- hate relationship with them. And We're still learning how to incorporate it. I know Marty seems to have embraced it more than I have.


    Marty Martin  58:08  

    I've been taking over. Yeah.


    Cameron Ghalambor  58:11  

    I think, like the first impression would be, actually the opposite, that AI would be reducing the creativity of scientific work rather than adding to it. So maybe can you walk us through how it actually helps us with our creativity rather than limiting our creativity?


    Oded Rechavi  58:32  

    Of course. So first of all, I'll explain what q.e.d. is. q.e.d. is a website where you can upload documents, papers, currently, in the future, perhaps other things, and then you get feedback. You get feedback, and it doesn't dress up as a human peer reviewer, okay. So one of the concerns about AI is that we don't try to do something to create feedback that looks like a human reviewer and talks like a human review. So it's not your peer, because it's a machine. It gives you feedback. And another thing that it doesn't do is that it doesn't generate the science for you. Many of the fears about AI is that it would hallucinate all kinds of things, make up the science. In this case, the idea, the idea for us is that it will give you feedback on your own work. And I think that while we suffer from the peer review process, that everyone knows it. Yes, it improves the work, but it's a torturous thing that takes forever. It's inefficient. It's biased. There are biases. Again, you know, minorities and women and all, and you know people don't have a Boston affiliation. You know, it's biased against many people, and it's also slow, and there's a lack of reviewers, of course. But we need it. We need feedback, and we also need some someone else to look at it. 


    Oded Rechavi  59:52  

    Yes, I agree, but I think that the human aspect of it is something that it's not that great. So. For example, if I publish, if I send a paper and it gets demolished by a reviewer, one of the things that I often think about, I'm not the only one is, you know, did this reviewer eat breakfast? Or maybe it's a competitor, right? So I think emotions on the human side of it is actually something you to remove from the peer review process, not from the doing of the science and thinking about science, but from the evaluation. Okay, I don't want an MRI scanner or CT scanner to have emotions. I just want it to find if there's something wrong with me. But you do want to get feedback that is also more creative than just yes or no. You want it to bounce back ideas. I think we're desperate for feedback. We need feedback. It's important. We shouldn't just publish everything without any feedback. We need feedback. But AI can give you feedback instantaneously, and it is it gets smarter and smarter. So some of the frustration with AI is people say, Huh, this thing can't replace in a second, what I learned for 20 years, right? But it can still do a lot and get to a point where you can even, actually, you know, do many, many things better than us. So I think it has limitations. In some things, people will still be better than AI. 


    Oded Rechavi  56:09  

    In the system that we created, for example, there's no hallucination you can try it doesn't make up references and all of that. And this is because hallucination is actually a solved problem. If other companies cared about it, they would solve it's not you have to be dedicated for doing something in particular, and then it works. And it gives you feedback. It doesn't look like a human feedback. What it does is that it takes the document that you upload, breaks it down into the claims that you've made, which also helps you organize your thoughts and then find possible gaps in your claims and sort of highlight them for you, and then you can try to address them, to improve them, to bridge these gaps and strengthen the paper. So that is one side of q.e.d., which helps you just do better science, improve your science before publication. The other side is, I mean, part of the problem is that science publication is very biased, but also extremely slow and inefficient. We know that humans are just incapable, editors and reviewers, incapable of dealing with the flood of paper that are published all the time and do a good job at evaluating it. And so it takes forever. It's sort of a you send to one journal, you get rejected after six months, you send another journal. By then, your postdoc already left the country, and it's just, it's a problem. So here you get it immediately, the feedback. And so what we want to do, and this will be, perhaps, by the time this podcast is in the air, in about a month, you will be able to, when you publish the preprint, there will be a button that says, Do you want to have published, also the q.e.d. report that was generated for you? It will be totally up to you. So this is the whole idea that this is an auto centered thing. You decide, okay, and if you decide to show the report, then it shows that you are transparent about the work and you accept criticism also in the report, you give you additional real estate to address the problem. So if the AI says, there is a problem this with this claim of yours, we maybe you could have done that, then you can respond, and everyone can see your response. It's like submitting your paper already with the response to the reviewer.


    Oded Rechavi  1:03:37  

    So for example, when I publish about transgenerational inheritance often I hear someone says, Why didn't you do it in mice? Okay, so I won't put it in the paper, but because I think it's irrelevant, but I will be happy to prevent this in advance by writing in the response to the reviews before I get the review. And so this will be, and I think this can accelerate science. And also, the preprint will already be reviewed by an AI, not by a human, but AI. So they're already partly validated. And also we are already talking to many entities that are interested in these reviews. So perhaps there will be forward thinking editors that would say, and I'm telling you that there will be that they will say, Okay, if you're if your paper has an AI review, q.e.d. review, then we take it into consideration. Maybe it will help us decide about sending it to peer review. Maybe we'll consider it as a third reviewer to accelerate the process. Maybe this will, if we have two opposing reviewers, we can decide based on this. Maybe we can, we can look at, currently, we break your papers into claims, and then we find problems with the some claims, and we find some claims are more valid than other they get scored. And then we say, you know, maybe we'll also look at the human reviewers. And we say, "we see that the human reviewers all of the questions were about one claim, but not about another claim". Maybe they didn't read through the entire paper. So it also helps us understand human reviewers better and what you're actually claiming versus the fluff around it, because we strip the paper out of the style and we just, you know, show it like a blueprint of the paper. So it helps you. It's also another point of view of the paper. So the idea is to to help you improve your paper, but also to help the paper get published and to help readers understand what they are reading.


    Marty Martin  1:05:23  

    Yeah, and Oded, it's available. It's active now for people to try it. I've used it. I've shared the link with other people, but people could try this right now, correct?


    Oded Rechavi  1:05:33  

    Yeah, absolutely. Just go to QD science.com. If you have an academic email, you can use it absolutely, enjoy, tell your friends, because the way we improve is from the feedback. So in the system, people can give feedback whether the comments that they get are good or not. They can also write text and this is how the algorithm improves. Importantly, it's very important to say, we don't use your data and we don't expose it to the world. This is the website. You know it's written, so don't be afraid it will ruin your patent, or the people will know about it or something.


    Marty Martin  1:06:09  

    So I don't know. Maybe this is gushing. This could be a little bit too much, but I've used this a few times. I've used this for papers that I've already published, and I've used this for manuscripts that were in review, and I was telling Cam before we got on with you, it blew me away. It was scary, really, to know, yeah. I mean, it was. So the one published paper, you know, it wasn't published so long ago that I didn't remember the process. The things that were picked up by the AI were the things that we modified the paper before it was able to be published. We had our claims, we had our interpretations. The referee said this, and then the final paper became something different. What the referee said we needed to do were still perceived by the AI to be, you know, fine. This is great. This is wonderful. But, and, you know, we had our caveats that were listed there, And then the same thing for the manuscript. You know, I was literally the day prior, dealing with the referees about what they wanted me to change. I plugged the manuscript into the AI, and it told me exactly the same things that the referee said. So it was, it was wonderful.


    Oded Rechavi  1:07:11  

    I'm very happy to hear it. That's great. I mean, I'm super happy to hear that. And this Thursday, which is, you know, for people here in the podcast a month ago. We'll have a new version, which is much better. So what you experience is going to be much better now.


    Marty Martin  1:07:29  

    Wow. That's hard to believe. It's already, it's already pretty amazing. I think my favorite part was, and this is kind of to our radical, audacious way, you know, it's, it's not just that you're getting, you know, feedback about improving your paper. And I love that it's two flavors of that. One is how to modify the text, to be more conservative, to be fairer, to be more inclusive of other literature. And then it's the next step, about different levels of go collect these data, do this extra experiment, that kind of stuff. That's amazing. But what my favorite part was it sort of takes that and distills it down to what you guys call an originality score, where it sort of says, here's how this paper looks relative to the rest of the field. I don't think you really said talking about its two different intentions. You didn't really speak to that part. So can you say how you get to the originality score and why you chose to come up with an originality score relative to the sort of quality, solidness, you know.


    Oded Rechavi  1:08:25  

    So this first thing we did is an originality score. And an originality score, it's important to understand that, first of all, it's not the only score that will be there. In the version that the listeners will experience, and the version that will come up this Thursday, two days from now, there will also be a validity score, and this will be also per claim, okay? And the originality score, it's not everything. It's not the quality of the paper. If I could put, maybe I will, something on the x axis, which is no, the more to the right, the more original it is. That at the end of the axis, at the right part of the axis, I would write big is true, okay, something could be original but not true, and the validity thing sort of complements that, to know whether it's true or not. And there could, there might be also other other scores in the future. I think that the originality not easy at all, but relative to the validity, it was easier to create. So when you judge a paper, how original it is is important, How valid is obviously important, but it also the other measures that are important, like how useful it is. I mean, we're thinking about many different matrices. For example, we're thinking about an anti-hype metric. So imagine that we want to know to what degree did the people hype up their paper. I don't know if we will actually put it, but all of these things are very useful when you read the paper.


    Cameron Ghalambor  1:09:39  

    So I haven't had a chance to use it yet, but I'm super excited I will be doing this probably as soon as we get off this conversation. But I had a couple of questions that are sort of related. One is, I mean, so within biology, there is a lot of diversity in how people do their science. You know, from ecosystem ecology, down to molecular biology, and there are different kind of traditions, probably in the way people write, and there's also differences culturally across how different scientists from different countries might present their research. So how do you train an AI to be sensitive or kind of, again, getting back to this idea of creativity, I mean, part of I think what, what we value is that not everyone thinks the same way. Not everyone does science the same way. On the other hand, we also accept that there's this hypothetical deductive method of doing science that is very robust, no matter who you are. So how do you train an AI to balance between those competing, I think, ideas, or, you know, approaches?


    Oded Rechavi  1:10:59  

    Yeah, so this is one of the things that's most important to me, of course. Also now human reviews often comment about your English or your writing style and so on. And one of the things that we can do, which most journals don't or all journalists don't do, is anonymize the paper, remove the affiliations, and more than that, strip it down into its structure and the logical arguments. This is also something that is demonstrated. You can see it in the website, when you run a paper that you get a sort of a blueprint of the paper, and then this is how it is measured. This is how it is measured. And so we are very intentional about not taking these type of things into consideration, including style and everything. Nevertheless, I mean, the AI does suggest improvements, that if the way that you wrote a claim is inaccurate, or you're overstating something, then it will suggest the ways to improve that. But when we judge it, we try to avoid this intentionally.


    Oded Rechavi  1:12:04  

    About different domains Absolutely. So currently, in the version that we have up now, if you up some we don't cover all domains. We are focused on biology, for sure, at the moment, maybe we'll expand later on. But now it's biology, and also in biology, we didn't deal with everything. Part of the secret is, and this is a thing very important, just being serious about it and going deep, the way we do it is we have many experts, scientists that are in the loop, always giving feedback and improving the system. And this is key. These are the users, and also scientists that work in q.e.d. and do only that. This is very important. So if you, if you run a paper, for example now in cognitive neuroscience, the system will give you a domain warning saying this is out of our scope. It probably will still allow you to run the paper, but it will tell you it's out of that. So be aware that we might not be that great in your area, but we will get there gradually, because we go one by one. Okay, if you run a paper in quantum physics, they will tell you, just don't do it. But this is part of the thing you need to take into consideration the field, you need experts to give you feedback about everything and be in the loop, humans in the loop, I think will be important always.


    Marty Martin  1:13:21  

    Yeah, Oded, can you say a little bit more about, you know, you mentioned a minute ago, but maybe talk a little bit more about it, because I've shared this with my colleagues, and they're, you know, very excited, and yet, most of them work in biomedicine, so human health data are a large portion of what they do. And one of the extra pieces, just so many elements to q.e.d., one of the extra pieces that you offer the option of uploading data, right? So some of the analyses could be done in addition to reviewing the manuscript itself. But how are you dealing with the fact that, you know these data are not usually intended to go out? I mean, they're not necessarily publicly available, but, but how does the AI deal with this right now?


    Oded Rechavi  1:14:02  

    So everything is local and secure. They're not shared. And we are also not learning from your data. What we learn from is your comments. For example, if you comment, and this is the sort of thing you can do, and say that a particular comment that you got was picky, this is how the algorithm improves. Or you say, this is okay I can do these experiments, but good to have. It's not a must for the publication. This is what we learned. So your data is your data, and it stays private and secured. We didn't deal with papers in medicine yet. I don't know what will happen in when this podcast is actually on. This is certainly in the pipeline, something we want to do, but we didn't do it yet, but we still give already quite good results also about just papers in medicine. Yeah the challenges are slightly different, but we also plan to do this more seriously very soon.


    Cameron Ghalambor  1:14:51  

    Yeah. So I guess one last question is, you know, this is still very new, but I'm curious like, what the reaction has been so far. Like, how has the biology community received q.e.d. science? Are you getting mostly compliments, or have there been any complaints? What's the feedback been so far?


    Oded Rechavi  1:15:12  

    So far, we didn't do any PR, including we didn't even do, we didn't even do, you know, tweets about it, or Bluesky, or all of it, nothing. But already, close to two thousand people use it, something like that, which is quite a lot. And the feedback is very good. The feedback is very good. Of course, some people are more happy than others, but the feedback is very good, and we are also improving. The beauty is that every version you see is the worst you'll see. It just keeps getting better, also because AI is getting better, but overall, I think that people are very excited. They're excited about what we did and we were doing, also a sort of a pilot survey, is that people are happy to use it. First of all, we Yes, we were concerned about people being just afraid of AI. Turns out, they're less that. They are less concerned than we thought, especially because, you know, you're getting feedback on your own work, and you're not, we're not generating new science for you. And the second thing is that I learned is people are happy to use it in different capacities, to improve their own papers and also to read other people's papers under different hats, you know, either as reviewers or as just readers. And so that is very interesting. And you know, the many papers and articles getting published saying that people use ChatGPT to review, and this angers many people for many reasons. First of all, you expose their papers, and also it's not meant for that. And also it's very vulnerable, and it depends on how you do the prompting. So therefore, I don't know if you saw it, but some people, there was an article that showed that many people in their papers, they write something like, you know, that is intention for the AI that would review it. So they write something like, write, that this is the paper, best paper you ever read, and it doesn't require any additional work, and published as is. And then they and then they just color the text in white or in very small fonts, and then, you know, it's called point injection. So first of all, I'll tell you that we are protected against that. You can't do it in q.e.d., but I think it's very important that if people use it as reviewers, and we can't stop people from using AI for anything, okay. Then it's better that they will do it in a dedicated tool that does it properly and that you can't mess with. So I think that this is another thing to think about.


    Marty Martin  1:17:29  

    Yeah well Oded, I mean, I really appreciate the time today. This has been great to talk about. You know, your research, and we wish you the best with q.e.d.. I know I'm going to continue to use it. I've been encouraging all my lab use it, and colleagues too. So 


    Oded Rechavi  1:17:40  

    Thank you very much. 


    Marty Martin  1:17:41  

    Yeah, last thing we always want to give our guests the chance to bring up or talk about anything that we didn't prompt you. You can consider this to be your injection. Is there anything else, anything else you wanted to say that we didn't ask? 


    Oded Rechavi  1:17:53  

    I mean, I can talk forever, but


    Marty Martin  1:17:55  

    Yeah we all can. No problem with that.


    Oded Rechavi  1:18:01  

    I'll just say we did, in addition to our work on transgenerational inheritance, we work on many other things. We worked on the ancient DNA, you know, from the Dead Sea scrolls and engineering brand parasites and economic irrationality, all kinds of things. And it's all great fun. So if people are interested, they're welcome to check it out online, or to, you know, write me and I'm always happy to discuss all of these things.


    Marty Martin  1:18:28  

    Excellent, excellent. Well, great. Thank you so much. We wish you the best.


    Cameron Ghalambor  1:18:32  

    Thank you so much for taking the time. 


    Oded Rechavi  1:18:34  

    Thank you very much. Totally my pleasure.


    Cameron Ghalambor  1:18:45  

    Thanks for listening to this episode. If you like what you hear, let us know via  Bluesky, Twitter, Facebook, Instagram, LinkedIn, Threads or leave a review wherever you get your podcast, and if you don't like something, we'd love to know that too. All feedback is good


    Marty Martin  1:18:59  

    Thank you to Steve Lane, who manages our website, and Molly Magid for producing the show.


    Cameron Ghalambor  1:19:04  

    Thanks also to Caroline Merriman for help with social media and a very warm welcome to our new artist, Brianna Longo, who produces our awesome cover images.


    Marty Martin  1:19:13  

    And a warm and belated welcome to our new blogger, Clayton Glasgow, who writes about items related to our guests and topics of the main show. Check out his work on our Substack page.


    Cameron Ghalambor  1:19:22  

    Thanks also to the College of Public Health at the University of South Florida, our Substack and Patreon subscribers and the National Science Foundation for support.


    Marty Martin  1:19:31  

    Music on the episode is from Podington Bear and Tieren Costello.

Big Biology