Discover innovative insights and highlights from the Curiosity Podcast by following us on Instagram and TikTok.
You can also listen and subscribe to Curiosity on YouTube, Spotify, and Apple Podcasts.
Transcription of our conversation with Hannu Rajaniemi, Founder and CEO of Helix Nanotechnologies
Immad Akhund (00:00:50):
Welcome to the Curiosity Podcast, where we go deep on a wide variety of technical topics with the smartest leaders in the world. I'm Immad Akhund, the Co-Founder and CEO of Mercury.
Rajat Suri (00:00:59):
And I'm Raj Suri. I'm the Co-Founder of Lima, Presto and Lyft. And today we're talking to Hannu, who is the founder and CEO I believe of Helix Nano. He is kind of a biotech savant. Helix Nano is focused on mRNA, and this guy seems to know everything there is to know about mRNA and it's past, it's present, his future, and he's working on some really interesting applications of this groundbreaking technology. What are you interested to talk to Hannu about?
Immad Akhund (00:01:29):
Hannu is super thoughtful. I think the interesting thing about Helix Nano is they're not just thinking about one application of mRNA, they're really thinking about it as a platform. So he's really gone deep in trying to think about how can you apply this to lab-grown meat and cancers and how did the COVID-19 vaccines work? So he's just got this broad set of thinking around it and he can also tie it back to how will this affect our lives and the social aspect, which I think makes it not just biology, but really brings it to life.
Rajat Suri (00:02:02):
Yeah, I mean, he's also a science fiction writer. He actually thinks about how biotech can affect our lives and he has really interesting comments on bio weapons and things like that, which I never thought about before. So really fascinating conversation and also just learned a lot about the science. I mean, the thing is with mRNA, I thought was we all kind of like, oh yeah, this technology came out, it delivered the Covid vaccine, but we've all kind of moved on with our lives, but the technology just gets better and better and there's going to be a lot more impact because of this technology, and he makes a great case for that.
Immad Akhund (00:02:36):
Yeah, it really actually made me think about the hype cycle of things. There was this crazy mRNA hype cycle, but actually it's from 1978 and these companies are already decades old. The next time something happens, maybe breast cancer is cured, or we have a preventative vaccine for everyone. It'll be like, oh, of course mRNA. And then there'll be another hype cycle about it. But it's great that people are continuously working on this in the background and we get to be the benefactors every 10 years or so of these breakthroughs.
Rajat Suri (00:03:08):
Yeah, I mean it's really compelling when he calls mRNA the transistor of biology basically,
Immad Akhund (00:03:13):
Right? Oh yeah, these guys have raised 60 million, right? So he's obviously a great storyteller on how the company can be hugely impactful.
Rajat Suri (00:03:23):
Absolutely. With that, well, let's welcome,
Immad Akhund (00:03:25):
Welcome, Hannu, thanks for being here.
Hannu Rajaniemi (00:03:27):
Thank you. Great to be here.
Immad Akhund (00:03:29):
Yeah. We wanted to kick off with how did you go from mathematics to building mRNA biotech company?
Hannu Rajaniemi (00:03:36):
This may be useful to start by explaining why I got into mathematics or mathematical physics, and that really I think has a lot to do with growing up in a small town in then also Finland, where winters are cold and long and very dark. But that does mean that when you walk home from school, you see a spectacular night sky full of stars. So I became an early space and sci-fi enthusiast. I really, really wanted to explore the universe and build spaceships, and that led me to study physics. What then pulled me fully in was this mystery of why does mathematics actually describe reality? So all our models of physics are really mathematical theories where often some very abstract branch of mathematics that has been developed completely independently turns out to be relevant for describing reality like differential geometry for general relativity, so gravitation or theory of hilbert spaces or group theory for quantum mechanics.
(00:04:32):
So I fell in love with that mystery, which still I think remains, remains a very deep mystery about the universe. And I ended up doing a PhD in string theory, which is of course an attempt to build a unified theory of all fundamental forces including gravity, which is something we have been unable to do so far. And there's other attempts to do that too, and there's sort of experimental window for string theory closing pretty rapidly, but it's a beautiful theory and unifies a lot of beautiful ideas, but it also didn't immediately lead to building faster than life flight spaceships. So I got a little frustrated with it together with another, similarly, this illusion strength here is Sam Halliday. We started a previous company which was essentially a consultancy solving problems in mathematics for industry. So applied mathematics company and we built a team of about 10 PhDs, mathematicians, theoretical physicists to work with us and then engaged with all kinds of clients, typically with some sort of hard engineering problem ranging from thermal modeling for satellites for European space agency to reservoir modeling for oil and gas, video compression algorithms for drones for aerospace and defense, and more speculative work around social network analysis, early explainable AI work, even quantum computing for the UK Ministry of Defense and actually also life sciences.
(00:05:47):
So we did do some work on optimizing, funnily enough delivery of drugs into so-called Langer hand cells under the skin, which actually have a lot of relevance to vaccines I obviously ended up working on later. But the biology was sort of a theme in the sense that it represents these complex systems that we don't really have the tools to understand well, in the same way that we have very fundamental mathematical theories in physics. So that already had some pull. There is this history or tradition almost of people coming from physics and mathematics going into biology. There's the RNA type club, which was this group of scientists in the sixties who really sold the genetic code. Essentially, what is the code for the instructions that RNA in the cell provides to the cell to make proteins? That club included a lot of physicists who had been also involved in the Manhattan project like Leo Ard and George Gamo, a famous cosmologist.
(00:06:42):
So there is a precedent for being naturally inclined to go from physics to biology, but for me, actually the reasons were also quite personal. That consultancy business ended up not working out really well. I think the issue that we came up against was not having a lot of confidence in ourselves with Sam as business leaders coming straight from academia. So we brought on board an external CEO, and that person was quite effective, but also insisted on bringing their life partner on board as a board director with a big equity stake. They essentially took over the company or we sort of got involved in a conflict over the direction of the company. It became a very toxic environment. I still hate checking email because I have email apnea, so I get shortness of breath from checking email that sort of dates back to that period. So in the end, we basically walked away from the company.
(00:07:29):
I then started to think about what to do next, and my mom had just been diagnosed with metastatic breast cancer. So she had had primary breast tumor 11 years previously, and that had been operated or resected. Then it came back and metastatic size to her bones. That obviously was quite a dark time and a dark, dark moment, and she eventually did not make it, but it did make me reflect deeply on the nature of biology. And then as I sort of looked more into cutting edge cancer therapies, it also became clear that there was this ongoing explosion of capabilities in both reading and writing DNA, where you sort of had these mors lawlike curves sometimes called Carlson curves after Rob Carlson who sort of first plotted and analyzed them of exponential reductions in the cost of reading a base pair of DNA or writing a base pair of DNA.
(00:08:22):
So it really felt very clear that if you project these trends forward, something really incredible is going to happen and it's going to fundamentally change not only medicine, but really what being human actually means. And then there was some other early signals that sort of showed glimpses of what that future might be. And one little gateway drug or a trigger for me actually was I spent time in Silicon Valley first in a place called Single Art University, and through there met this Israeli scientist called Ida Bait who had built little DNA nanorobots. So DNA can actually be used not only as a recipe for making proteins, but for as a construction material. So there's this whole field called DNA origami where DNA structures are sort of folded into different geometric shapes and they can be then assembled to do functional things. So Edo had built little nanorobots, these very simple mechanical devices out of DNA origami, so like boxes that can contain drugs that open when they encounter a certain kind of molecule.
(00:09:24):
And this has blew my mind. That whole technological direction has not really panned out yet, at least as a useful set of tools, but just as a demonstration of what was possible, that to me indicated that the future was going to be spectacularly weird, and it was going to be kind of almost like Eric Drexler style nanotechnology was actually closer than I had imagined, and it was really biology. So I decided to fully jump into biology. It had that sort of same feel that I had gotten from this deep connection between mathematics and physics, that there was this whole new universe that had evolved through billions of years full of wonder and complexity and something that we were now learning to talk to and find tools from and repurpose.
Immad Akhund (00:10:09):
One thing beautiful about physics is that it can be modeled with mathematics, and the mapping seems fairly strong, whereas at least my impression of biology is that it's way messier. There isn't clear mathematical modeling and determinism. Maybe is mathematics a good way of modeling biology? And at least at the micro level you guys do.
Hannu Rajaniemi (00:10:33):
There will undoubtedly be mathematical principles that will be applicable to biology. I think our expectations should be different though than from physics because we are talking about these complex emergent phenomena. But I think we will maybe with the help of AI understand the principles of emergence and how these complex structures emerge from simpler structures. There is some interesting work already in that direction. For example, I think my favorite example is there's an MIT scientist called Jeremy England who has been working on non-equilibrium thermodynamics to try to build essentially toy models of life. He has noticed, for example, that this emergence of biological structures like cells, et cetera, can actually be explained if you define this quantity of dissipation, like maximizing how much energy that system dissipates and actually self replication emerges very naturally from that. So he primarily works with very, very simplified models of life-like systems and biological molecules.
(00:11:31):
But I think fundamentally it is going to be possible. Now, there's also some very clear, not necessarily mathematical, but there are also clear organizing principles that exist in biology. One really important one is that biology is an informational field. Cells are information processing systems, and our immune system, for example, is an information processing system. So we can look at and manipulate the information flow in that system. We can also intervene in it without fully understanding it, we can at least poke at it and perturb it. That's where the power of mRNA lies actually, that it is the natural layer where to intervene in that information flow.
Rajat Suri (00:12:09):
These cells or these molecules being information processing systems, is it possible to write really good simulations of how all these different systems interact and be able to predict accurately how these different systems will act in different situations?
Hannu Rajaniemi (00:12:25):
Not so far. I mean, I think the complexity of an individual cell is really quite staggering. I mean, you have many, many thousands of different kinds of proteins, each of which is like a complex nano machine in its own right with millions of copies of them all crowded together in this bioreactor mess, which then obviously interacts with its environment and other cells constantly as well. So some progress is being made. There's this nice quote from Demi Hassabis from DeepMind, founder of DeepMind around drawing a parallel between physics and mathematics and then biology and ai. I think machine learning does provide us with a natural way of approximating very complex non-linear systems. If we can generate a lot of data about them, I'll need to check the exact reference, but I think the most impressive thing in the kind of whole cell simulation direction I've seen is a deep learning model of a yeast cell, like a single yeast cell that does seem to be able to at least to some extent, predict gene expression patterns in an individual yeast cell.
Rajat Suri (00:13:26):
What's the gap then? I mean it's obviously complex, but we have amazing computing power at this point. Is it that we don't have an accurate model for how these things react or because it doesn't surprise me at this stage where we're able to do things like LLMs, we can't predict biological interactions as well.
Hannu Rajaniemi (00:13:44):
So there's certainly been amazing advances in terms of protein structure modeling, but I think the dynamics is still very challenging. We can kind of predict the static 3D shape or a protein, but then actually the sort of how it changes dynamically even for an individual protein is very, very challenging. And I think that, I mean, it might be that we need quantum computers. It is ultimately a quantum mechanical problem. What is the landscape of dynamics of a protein molecule or another kind of biological molecule look like? And that's where I think we run up against the limits of classical computers. Classical simulation of quantum systems becomes exponentially, exponentially hard with size. So to flip it around, the system itself simulates itself rather than sort of trying to necessarily fully build simulations of cells or other biological systems, I think we just should also just develop better ways of interrogating them, getting data from the actual systems themselves because this is actually why drug discovery is so hard or has traditionally been so hard.
(00:14:43):
If you look at the cost and failure rate of drug development. So for cancer drugs from the first phase one safety first safety study to approval, the success rate is 5% and it costs a billion dollars of that order of magnitude. So clearly we are doing something wrong, and I think one of the fundamental things we are doing wrong is that we don't have very good predictive models in cancer. Like the most successful therapies are the ones that engage the immune system to go after cancer. Mouse immune systems and human immune systems are not the same. There are sort of tumor organoid models where you take part of the patient's tumor and culture it with human immune cells, but then you don't have a functional immune system that actually develops and learns a new response computationally. We have also not at all yet bridge that gap because of all the sort of scaling issues we just talked about to predict like an immunotherapy efficacy.
(00:15:34):
It's not that you need to model one cell, which we cannot do, but you need to model a system of billions of interacting immune cells and pathways between them. And then the cancer itself where you have something as complex as cancer is our own cells. So it has that full genetic complexity that we have. So there's some humility there that I think we need to have around how much can we assimilate, but the system itself simulates. So I think if we can gather more data from the system, which we now can with both high throughput, DNA sequencing and other new high throughput assays, we can generate large datasets that we can exploit. And then if there's a natural way to perturb the system, like with mRNA, I think we can close the loop. And then Demi Hass point, given the feedback loop, we can then have a AI design layer that helps us actually steer the system in the right direction. So it might be also more like a real-time control problem rather than fully modeling the system.
Rajat Suri (00:16:34):
So using the outputs of the system, using the information we get to train the inputs, understand the system better, understand that black box similar to ai.
Hannu Rajaniemi (00:16:43):
But to make that sort of not abstract here is something that mRNA makes possible. Now, this hasn't really been done, but it's possible. So let's say we have a cancer patient, we look at their tumor and we find that there are some immune cells that are finding their way into the tumor and are sort of successfully attacking the cancer. We can now sequence the genome of those immune cells. We actually can learn what is it that they are targeting. So immune cell, like a B cell, each of them makes a different antibody essentially that binds to different things, and that is genetically encoded. So we can read that out. So we can read out what antibody is an immune cell, B cell making itself its way into a tumor. We can immediately take that sequence and turn it into mRNA and put it back into that patient to amplify that response.
(00:17:39):
So to amplify their own existing response, their immune system has already figured out what is the right kind of response. It's maybe just not powerful enough. So you can amplify that and then you can also put that into another patient. Another version of this, which is sort of more clinically later stage now, is the idea of personalized vaccines. This may have some fundamental problems as an approach, but what is clinically being done now not yet approved as a therapy, but s like Moderna and BioNTech are taking individual patients tumors, sequencing them and identifying mutations that are unique to those cancer cells and then making vaccines in an attempt to target those mutations. These haven't yet moved the clinical needle massively because often it is actually one problem with cancer cells is that they are our own own cells, so it's hard to generate immune responses against them and also specific individual mutations. Cancers can usually evolve around these kinds of feedback loops are now possible. In the case of these personalized, so-called neoantigen vaccines, there is already machine learning steps the sort of in the process of deploying these therapies, machine learning models are being used to predict which ones of these mutations might actually be visible to the immune system and sort of displayed on the surface of the cancer cell so that the immune system can attack them.
Immad Akhund (00:18:53):
Maybe this is a good point to talk about what does Helix nano do, how long have you been working on it, and what was the idea and the evolution and what does it do now?
Hannu Rajaniemi (00:19:03):
So basically the earlier origin story really came directly from my excitement around biology. I sort of wanted to find some key critical problem to solve solving biology, and actually the original idea had to do with being able to record what happened inside an individual cell and using DNA as sort of information stories for that. So kind of a crazy idea now, actually quite a lot of work in that area. But that was a good enough idea for a business plan competition. Around 2015, the business plan competition was organized by Johnson and Johnson, and the winners basically were given JJ support and also lab space at the j and j campus in Belgium. Janssen Pharmaceutical ended up spending a year at Janssen working on this idea with a couple of scientists, but also talking to everyone involved in the drug discovery process. So that was a big part of my biology education of just trying to understand what was happening and trying to identify where some of the critical bottlenecks there where.
(00:19:59):
And what very quickly emerged was that sure enough, it is now possible to print genes and make synthetic DNA and design DNA constructs, but how do you actually get them to do something useful in the body? Getting them to actually then work in the body was sort of a big challenge. And I made a connection with the resident university scientist who had come from George Church's lab at Harvard Medical School. So George Church of course is the sort of Titanic figures in the field of synthetic biology whose lab has invented both key DNA sequencing and DNA synthesis technologies was very early involved in CRISPR and many other things. Now also operating very much in this sort of biology AI intersection. So the Dresden scientist introduced me to Nikolai now my co-founder and the CSO of Felix Nano, and together with George and Uri, Nikolai essentially some of the core modern high throughput DNA synthesis technologies, how to use inkjet printers to make smaller strands of DNA that could then be stitched together for actual genes.
(00:20:58):
So his work went into a company called Gen nine that Ginkgo Bioworks bought, and then Nikolai stayed on at the church lab to try to figure out how to engineer the next generation CRISPR like systems, but also came up against this problem of how to deliver them, how to get them into the body. And that's where we found a meeting of minds and concluded that there's sort of a finite space of solutions. It's maybe worth just unpacking a little bit of what happens in the cell, what is the information flow that we talked about? And our DNA, our genome is where the information is more permanently stored. So it's the recipe book for all the proteins. Proteins really do everything, but the recipes for the proteins are stored in DNA. When a cell needs to make a protein, it makes a temporary copy of the recipe from the DNA into mRNA.
(00:21:44):
DNA is the rom and RNA is the ram. The program gets loaded into memory and RNA is sort of less permanent chemically, it's more transient, it's less stable chemically than DNA, and that's why it is on purpose. The cell doesn't want to make every single protein permanently all the time, but only when it needs them. And then the RNA carries the information to the protein factories of the cell, the molecular machines that actually then make proteins and they're called ribosomes, and they are actually very much like 3D printers. They read through the mRNA and then one amino acid by amino acid. The building blocks up the proteins, they assemble the protein, and then the protein folds into its final 3D shape or can also be dynamic shape that then goes on, carries out its function, whether it's construction material or signaling or enzyme or something else. But that is the inflammation flow. So D-N-A-R-N-A protein, that's called the central dogma of molecular biology.
Immad Akhund (00:22:35):
This RNA is called mRNA, right?
Hannu Rajaniemi (00:22:37):
There's other kinds of RNAs, but the type of RNA that carries the information to make protein is called messenger RNAs or MR A mRNA. We were initially more interested in DNA, so could we just get synthetic DNA more efficiently into the cells more directly? And that's why the company is called Helix. Nano Helix is the DNA helix. So one way to get DNA into cells, which is used a lot with gene therapy is with viruses. So you make a synthetic virus and you load the DNA in and infect the patient with the virus, and it's a non-replicating virus. You've removed all the parts that help it replicate. So you're just sort of leveraging the parts of the virus that help the virus get into the cells. And this kind of works. There's the class of viruses called AAVs or adeno viruses, and there's many, many, many gene therapies in clinical development using that technology.
(00:23:24):
But we sort of discarded that approach pretty much immediately because it has fundamental limitations. First one is you get an immune response because you're putting the DNA into a viral vehicle, a virus, your immune system, once it's been exposed to that virus once recognizes it and now you can't treat the patient again. So what happens here, the virus doesn't necessarily integrate the DNA permanently into your genome. It just puts an extra bit of DNA into the nucleus that sort of floats alongside your own genome. And then RNA gets made and protein gets made from it. So it does work, but that can also then get flushed out eventually. And then you can't ose because now you have an immune response against the virus. Incidentally, this is kind of an issue also with some of the viral vector vaccines that were used in covid vaccines like the adenoviral adenoviral vaccines like j and j and AstraZeneca.
(00:24:15):
The other problem is that viruses have a limited DNA payload capacity, so you can only put in so much DNA and it's quite small. So we thought viruses were out, we wanted to go fully synthetic, and we found a way to make these minimal DNA vectors like basically just take synthetic DNA generated using some of these technologies that Nikolai invented, put little loops at the end to make it more stable so that it doesn't unravel. And then the question is how do you get it into the cell? And then more importantly, how do you get it into the nucleus in the cell where DNA needs to be to work? And we realized actually that there was a way to solve that nuclear import problem of nuclear entry problem. There's various proteins that have signals that tell the cell to get them into the nucleus because there are proteins that need to go to the nucleus to do stuff like unwind DNA or make mRNA.
(00:25:03):
So we thought we would piggyback on that pathway. We designed a protein that had a portion that grabbed the DNA and then another portion that told the cell to take this whole thing into the nucleus. And that actually worked. It enhanced our ability to get DNA into the nucleus. But then the problem was we didn't want to actually separately make that protein. This is sort of, again, one of these assumptions that I've maybe had here in the background that's useful to unpack. Making proteins is hard. The first biotech revolution going back to the late seventies and Genentech and so on. The realization was that you can take a human gene and put it into other organisms and have other organisms make human proteins, so insulin or other biologic drugs. But the problem is that every protein is different. So to go through that process, you'll basically say have to go think, okay, in what kind of cell line should I do this?
(00:25:57):
Is the Chinese hamster ovarian cells? Is it yeast? Is it some other cell type? Where can I actually make this protein or what cell type can make this protein for me? And then the question is, okay, now it's sort of working. Can I purify the protein? Can I isolate enough of it with high enough quality? Does it sort of clump? Does it do something weird that I didn't need to deal with? And now it's six months later and you still haven't made your protein to scale that? That's a billion dollar exercise. Also, we did not want to do that. I mean, that's kind of one of the coolest things to keep in mind about working with DNA and RNA. This is about turning your own cells into these protein factories and having them do what they do naturally to make these proteins right there where they're needed.
(00:26:37):
So to go back to our two stage rocket system for getting DNA into the nucleus, we then realized that, okay, RNA doesn't need to get into the nucleus. RNA just needs to get into the cell. So what if we have RNA make our helper protein and we put both DNA and RNA and simultaneously and that sort of helps get the DNA into the nucleus. And that also worked. But then looking at our controls where we just used the RNA, the expression levels were so high that he actually felt like we didn't need the DNA at all. And then that's kind of going back to this information flow in the cell. The RNA layer is special. So RNA is not permanent. Unlike DNA, it has the same design freedom, so you can have the cell make any protein you want, not just any human protein. This is one of those sort.
(00:27:26):
Also galaxy brain moments. Our cells can make any protein that can exist if given the right information, not just the 20,000 human proteins, but I don't know what the number is, like 10 to the power of 1300 or something like absolutely ridiculous number of what is the number of possible proteins our cells can make? All of them if they're given the recipe and the RNA is the most direct way to give them that recipe. And this was like late 2017, early 2018, we were fully in on mRNA. It became a very defining belief for Helix Nano that mRNA was going to be the transistor of biotech. It was going to be the layer on which everything else would get built. This would be the place where we interface with human biology. That's kind of how we ended up with mRNA. And then the next question was, okay, if mRNA is the thing, why aren't we there already?
(00:28:15):
What are the fundamental problems in the field? What has prevented mRNA from taking off given that this is sort of a fairly obvious observation, that this is the right layer. And actually historically, the first experiments with synthetic mRNA expressing a protein in human cells go back to 1978. This is not a new idea, but it did take a long time for the field to take off because the problem is that if we're trying to do this, Manny, the middle attack on the cell by putting in mRNA from the outside, this is also what viruses do. So viruses try to play exactly the same trick to hijack our protein making factories to make more themselves. So the cells themselves, not just our adaptive immune system, but actually our cells themselves have ancient sensors that have evolved to detect RNA that looks viral. If we make mRNA the same way the cell makes it outside the body and put it in above a certain dose or a certain dosing frequency, these sensors inside the cell, these alarm bells warning the cell that it's been infected are going to be triggered.
(00:29:22):
And then what the cell does, the cell shuts down its protein factories. It doesn't want to make more virus, but that also means that your mRNA payload does not get made. And then furthermore, it alerts all the nearby cells. It sends out inflammatory signals to all the nearby cells. That effect can also become systemic. So if you had a Covid mRNA vaccine, some of the local side effects, some of the local inflammation actually come from this effect. So the reason we have covid vaccines, the reason the mRNA field exists is basically Katalin Kariko, Andrew Weissman who just got the Nobel Prize in medicine for this. The really interesting observation that Kako made was why does our own mRNA not trigger these antiviral sensors? If our cell is making mRNA all the time, why doesn't that mRNA set off all the same alarms? How can our cells even possibly work? So what she realized was that after the cell makes mRNA, it adds various chemical decorations to it to market as self as something that is actually us. And she found a way to duplicate those chemical modifications synthetically and to make mRNA that was actually chemically modified in a way that made it look less like a virus and more like our own RNA.
Immad Akhund (00:30:38):
When did she discover that?
Hannu Rajaniemi (00:30:39):
2015.
Immad Akhund (00:30:41):
What is that method called?
Hannu Rajaniemi (00:30:42):
So chemically modified nucleotides. So RNA, like DNA is made out of these four base pairs or four letters. So DNA is a CG and TRNA is a C, G and U. And the specific modification that she initially discovered with Weissman was called pseudo uridine. So you take the U and you add a chemical group to it to make it look slightly different. That was kind of the firing shot for the field.
Immad Akhund (00:31:06):
Didn't Moderna start before 2015 though, or is that around when it started?
Hannu Rajaniemi (00:31:11):
There's a pretty direct line from this discovery. Maybe I'm thinking the key publication was 2015, but the work was out there earlier. What happened was there was another scientist called Derrick Rossi whose lab was working on stem cells. So he was trying to figure out how to turn normal cells into stem cells to make induced pluripotent stem cells. This is of course the discovery of Shiri Yamanaka. There are these Yamanaka factors, which are genes that if you turn them on, you can restore cells to a stem cell-like state. So Rossi realized that mRNA was a really good way to deliver these Yamanaka factors, but then he couldn't make it work with normal mRNA and I forget now the name of the scientist who did this in his lab, but they realized that using these car co modifications that actually worked. So Rossi then realized that this had a lot of potential that really chemically modified mRNA could transform the stem cell field.
(00:32:03):
And I believe you had Bob Langer on the podcast recently. Rossi then went to Bob Langer to say, there's something interesting here. And Lger realized that no, no, this is much bigger than just stem cells. This can be applied to anything. And then Bob went to Novar afe on flagship pioneering, and that's how Moderna got started. But the discovery was, and actually the IP did come from Kariko and Weissman, so the pseudo uridine, which was the original Kariko and Weissman discovery turns out that's less useful if your mRNA manufacturing quality improves. But there is a derivative of that called N one Metals to the uridine, which is now the gold standard of the field, which she also discovered, and I think Moderna did discover that independently, but Kariko filed her IP first. So both Moderna and BioNTech, the other sort of big mRNA company who collaborated with Pfizer on the Covid vaccines licensed in one uridine from UPenn and Kariko.
(00:32:56):
There's a story there also around how Kariko was treated by UPenn. She was actually in the process of doing her own startup to develop mRNA therapeutics, but UPenn decided to take the IP back and then license it to Moderna and BioNTech. So she's somewhat redeemed now by the Novell Prize, but that is kind of the origin story of the modern version of the mRNA field. It is really all about how to make mRNA look less like a virus. We kind of also realized that even though anyone with Uridine enables higher doses and more frequent doses than unmodified mRNA sort of natural looking mRNA, it's still not perfect If you exceed a certain dose threshold, even with modified mRNA or certain dosing frequency, you still trigger these antiviral sensors. That fact you can see more or less directly from the pipelines of modern non BioNTech, they are kind of focused on mRNA applications where you don't need to dose that frequently or the individual doses don't have to be that high.
(00:33:56):
So that includes various types of vaccines. It includes single dose immunotherapies, it includes some very low dose requiring enzyme replacement therapies for rare genetic diseases. So the limitations of technology are built in to the kinds of things they're doing. So one of the core ideas behind Helix Nano is we are sort of constantly pushing the envelope of what is possible with mRNA. We want to turn mRNA into this almost like an external genome, like an exo genome for us to store outside the body and incorporate into our bodies whenever it's needed. And big part of that is making a larger and larger and more frequent mRNA doses possible. One core breakthrough we made is a novel chemistry that is analogous to the cardiac discovery but actually modifying two of the mRNA letters simultaneously, specifically C and U simultaneously. And we're finding that lets us go at least nine times higher than was possible with N Metals and to expand the space of possible mRNA applications. So that includes things like ultra potent vaccines, but also turning potent biologic drugs into mRNA.
Immad Akhund (00:35:01):
Are you actually building your own vaccines and other things?
Hannu Rajaniemi (00:35:06):
Yes.
Immad Akhund (00:35:06):
Okay. You're not selling this kind of IP slash methodology.
Hannu Rajaniemi (00:35:11):
Again, this is like a transistor level mRNA technology. It is so broadly applicable that we are not going to be able to capture all the value ourselves, nor should. We are also currently in a number of collaborations where we will license that technology to others for other applications that we're pursuing ourselves. But we are certainly building our own vaccines and therapeutics where we are aiming to go. The massive opportunity we see right now, which is sort of uniquely possible with mRNA, has to do with closing that loop that we talked about going from DNA sequence information that we can read from DNA to a drug or a therapeutic. And one setting where we think this is going to be incredibly powerful is preventative cancer vaccines or a very, very early stage treatment cancer vaccines, but effectively preventative cancer vaccines because the other set of technologies that has been advancing incredibly rapidly on the back of DNA sequencing that has followed these Moores Lawlike curves like going from hundreds of millions per human genome to a hundred dollars per human genome in the space of 20 years, one area leveraging that are liquid biopsies.
(00:36:15):
What that means is you can take a blood sample and little bits of DNA in that blood sample can be amplified and sequenced to detect any signs of early development of cancer. So pre-cancerous cells and cancer cells in your body are shedding these DNA sequences into your bloodstream and we now have the technology to detect them and then predict whether you might be developing cancer or not. There are companies like Grail, many others who are working on this, but what's actually missing is an intervention, let's say as a readout says you have X percent chance of developing pancreatic cancer or colorectal cancer in the next five years. You're not going to go on radiation or chemo at that point. You might not have any detectable tumors, but you have cancers or precancerous cells in your body and your immune system theoretically knows how to deal with them. It just needs a little bit of information. And in the form of mRNA vaccines, we now have a way of giving that to you. But to really enable that, we do need more potent mRNA vaccines that have been possible in the past.
Immad Akhund (00:37:17):
In this case, it would have to be personalized per individual or there will be classes of vaccines you can develop.
Hannu Rajaniemi (00:37:23):
There would very likely to be classes. So there would be some common patterns that would recur. In fact, that's kind of born out by some studies that are coming out. So they are common pathways that get activated in early cancer development that are broader that you could target. So it might not have to be uniquely personalized, but you could have sort of a library of things that you mix and match for a given individual.
Rajat Suri (00:37:44):
Where is it the furthest along? We are today in the state of that technology for developing sort of a cancer vaccine.
Hannu Rajaniemi (00:37:50):
So there are a lot of therapeutic cancer vaccines in development, both mRNA cancer vaccines and others where you are sort of trying to essentially amplify a cancer patient's response against an existing tumor. So those are sort of been in late stage clinical trials currently, and there is some promise modern has seen a response at least reducing the rate of recurrence for metastatic melanoma. BioNTech has seen extended survival in pancreatic cancer, which is really one of the most lethal cancers. So there is promise there. The challenge though is still the response rates, so the response rates are quite low.
Immad Akhund (00:38:25):
What do you mean by response rates?
Hannu Rajaniemi (00:38:26):
So how many patients actually have their tumor shrink?
Immad Akhund (00:38:29):
So it works, but it only works for a small percentage of people?
Hannu Rajaniemi (00:38:32):
Correct.
Rajat Suri (00:38:32):
What percentage are we talking?
Hannu Rajaniemi (00:38:34):
About this highly variable between different cancer types and trials, but it is between 10 and 50, but probably more on the lower end.
Rajat Suri (00:38:40):
And still in small scale, right?
Hannu Rajaniemi (00:38:42):
No, these are hundreds or around a thousand patients. Patients I think is now the phase three trial that Moderna and Merck are doing.
Rajat Suri (00:38:49):
So promising, but there's still some work to do.
Hannu Rajaniemi (00:38:51):
This is still also the first generation mRNA technologies. I think there's at least three layers at which these things can be improved. The amount of vaccine you can deliver, how strongly can you engage this immune system with that vaccine, and then what are you targeting the right things in the cancer? So I think there's order of magnitude gains to be made in each of those areas.
Rajat Suri (00:39:12):
Yeah, absolutely. You also mentioned on your website that looking at targeting climate change, how is mRNA going to have impact on climate change?
Hannu Rajaniemi (00:39:20):
I actually can't talk about that too much. There are at least two different applications. I can talk about one of them. One area that I think will have a big impact on climate change is food. So reducing the carbon footprint of food where a lot of people are excited about cultured meat, so actually growing meat in the lab, growing meat in bioreactors rather than slaughtering animals. And the cattle industry obviously has a massive both carbon footprint and land and water use footprint and so on. But the challenge has been scale. One big issue with cultured meat is that there are these certain components that are present in the serum like cow serum when you actually culture cells that you need to include that if you're not using any annual products, you then need to make synthetically. So you actually need to set up a cell powered biologics factory, make these components called growth factors to help your meat sales grow.
(00:40:11):
And that's a very, very expensive process. These growth factors are enormously costly to produce and to some extent that'll obviously get cheaper with scale. Still that is going to be one of the main bottlenecks for scale for cultured meat. So something we have explored and some experiments with partners have some IP around is just using mRNA to encode growth factors. So just to have the cultured meat cells make those growth factors themselves. And mRNA manufacturing as we've seen also with coded vaccines is extremely scalable and can be made extremely cheap. So that's one area where I do feel there's a lot of potential in a way actually very similar even to the Derrick Rossi idea behind the modernist story, like how to manipulate cells in cell culture and change the cell type and get them grow faster.
Rajat Suri (00:40:58):
Yeah, basically any biological process where you need to manipulate production, you could use mRNA to effectively do that, right?
Hannu Rajaniemi (00:41:06):
Yes, that's correct.
Rajat Suri (00:41:07):
And so you're focusing on these emerging areas that big markets, if you figure something out there, then you can come out with a product that could help a lot of people.
Hannu Rajaniemi (00:41:16):
That's correct, yeah.
Immad Akhund (00:41:18):
Yeah. The normal startup advice is focus and do one thing really well. It seems like you guys are doing a bunch of different things. Is there something about biotech that works better with an broad set of parallel experiments as they take a long time to get through trials and stuff?
Hannu Rajaniemi (00:41:33):
I would say we are converging around like 98% on cancer currently. We've certainly gone through quite a lot of exploration on the way it is. If you are an AI or a machine learning company that has built a better foundation model, what do you apply it to? I think you need to do a lot of experiments to figure out where is the biggest unmet need, the highest value problem, and the most impactful problem you can solve. So that's kind of the journey we've gone through and I think both in terms of our personal passions and the long-term impact, closing this information loop in oncology is what we really fundamentally want to do, but there are other areas where we can enable others with the technologies that we've built on the way. So I do agree on the focus point, but also one has to switch between the exploration and exploitation modes. There's a place for both.
Immad Akhund (00:42:17):
This is probably a little bit of a naive question I always imagined, especially as a kid that we would solve cancer, there'd be one solution and all cancer would be solved. Is cancer like that or is it always going to be kind of like a whack-a-mole where you'll be like, oh, this type of cancer we can now do, but there's another 10,000 to go.
Hannu Rajaniemi (00:42:36):
There might be a silver bullet, there's at least a chance. There are certain targets in cancer that are fairly universal. We are exploring some of them. We are developing a vaccine candidate targeted against one of them. So I think it might be a finite set of things, not necessarily just one thing, but there might be a finite set of things. Then I think the point is also at what point do you intervene? I mean I think more broadly immune system is the silver bullet. The immune system can cure cancer, clearly it does. So every day we have cancer cells or precancer cells constantly popping up in our bodies and we don't get cancer, all of us all the time. So it is successful at this.
Immad Akhund (00:43:13):
Does it fail because it's just a probability game? Eventually something gets through.
Hannu Rajaniemi (00:43:17):
It's a probability game and it is an evolutionary game. So the immune system puts evolutionary pressure on the precancer cells. There's a set of steps they go through to then hide from the immune system to proliferate faster. Under that pressure it evolves and then eventually it sort of reaches the complexity threshold where it can sort of have its runaway velocity where the immune system can't keep up. And at that point we obviously then need to help the immune system. I actually do think there is a finite set of universal targets in cancer that will be very powerful and hitting those targets is harder. One challenge there is that a lot of these universal cancer targets are also targets that would be present on healthy cells, although usually in lesser amounts. But the immune system has also evolved to avoid attacking ourselves. So that is one of the fundamental problems in cancer cancer is us and the immune system doesn't want to attack it. There are these mechanisms called tolerance where the immune system learns to avoid targets that are present on normal cells and a lot of these universal cancer cell targets are in that category, but there are ways to break that and something we've actually been able to do recently leveraging some LM based protein design tools to be able to overcome some of these self-tolerance effects. And we are now testing that.
Immad Akhund (00:44:32):
What does that mean?
Hannu Rajaniemi (00:44:33):
We are taking a target that is present on a cancer cell and also on a healthy cell. So if we just use that target by itself as a vaccine, we wouldn't generate a very strong immune response, but there's a way to redesign that target that we then deliver as mRNA in such a way that it sort of evades this self tolerance issue.
Immad Akhund (00:44:53):
Is this like a fine tuned LLM that understands proteins?
Hannu Rajaniemi (00:44:57):
So LLMs are actually incredibly good now at generating protein designs. So for example, there's this open source meta release model called ESM two that was trained with 350 million protein sequences. That has actually, the remarkable thing there is that actually, even though it was not told anything about the structure of these proteins, it was not given any structural information. Just in the process of compressing that information and learning how to represent these proteins internally, it actually came up with the representation of structure and is now outperforming alpha fault like systems in being able to predict protein structure and these kinds of models are also generative. You can essentially specify a set of properties that you want for your protein and you can prompt it, prompt it with that and have it generate designs for you. So that's what we are leveraging for cancer vaccine design.
Rajat Suri (00:45:47):
I mean you guys have been around since 2013, is that right?
Hannu Rajaniemi (00:45:49):
So this company was technically founded 2013 as an entity. I would say in this modern form we go back to 2016, so that's kind when I started. So I had this in the wilderness with a lot of help on the way early on, but then with Nikolai we really converged on the current form.
Rajat Suri (00:46:04):
And how is the business model? Are you basically, it's a technology driven company, so you're trying to one day come to, I'm guessing, so you're trying to develop a technology that one day that could be worth a lot, but it's kind of binary outcome. Is it either develop technology or you don't?
Hannu Rajaniemi (00:46:19):
We develop technology in order to develop products or let's say 20 years to be conservative. We want to have a product out there that gets cancer deaths in the US down to less than 10,000 a year.
Rajat Suri (00:46:29):
Wow. So 10 to 20 year horizon is kind of what your development horizon is and today you've raised a bunch of funding.
Hannu Rajaniemi (00:46:36):
We raised 48 million today, so that's sort of now enough to get us to transitioning to clinical stage, which is happening early next year. I mean this is kind of one of those open master plan like things. So unpack the three steps. How do we get to cancer prevention? How do we build the company to do that? So we've talked about some of the core technologies we've developed those. The core technologies are valuable and so we can generate some early revenue through licensing those technologies to others. As we then start developing actual cancer vaccines and other mRNA therapeutics, we actually don't necessarily have to take them all the way in the clinical development process, but once we get to phase one, phase two, we can license and partner them and then they are sort of much higher value. Those are much higher value deals than technology licensing deals. And that led then lets us bootstrap to a full stack biopharma company where we can run a very large five year, 40,000 patient cancer prevention trial. Then I think things become really, really interesting. So it is definitely a long-term game, but in terms of the impact, I think the timelines are actually pretty short.
Rajat Suri (00:47:35):
Sounds good. And so the game is basically in the short term develop patents and technology so you can license it out, which gives you the revenue you need?
Hannu Rajaniemi (00:47:43):
And products to get to and products and product and products.
Rajat Suri (00:47:44):
Okay. How big is the team?
Hannu Rajaniemi (00:47:47):
We're 20 full-time people, 20 researchers. And then we have various support organizations where let's come up to about 13 FTEs altogether primarily around chemistry and manufacturing.
Rajat Suri (00:47:59):
And you have a lab or some kind of physical space?
Hannu Rajaniemi (00:48:02):
We have 20,000 square foot labs space in Seaport in Boston.
Immad Akhund (00:48:06):
And a lot of people were very excited about mRNA covid vaccine, but it didn't feel like it delivered on the promise of giving you long-term immunity, I guess. Is that just something about covid, it just mutates too much and there was no vaccine that would do it or is it related to mRNA?
Hannu Rajaniemi (00:48:24):
It is definitely not a mRNA property. It is really all about the speed of Sarro COV two evolution, which essentially the challenge is that you have this virus that has sort of the part that it uses to enter our cells. The spike protein is very, and specifically the kind of tip of, it's called rrb D receptor binding remain is very big and very flexible, so it can easily tolerate a lot of variability. So it actually mutates more slowly than the flu, but more of the mutations are viable. A sort of bioinformatician scientist called Trevor Bedford has run this analysis of what is the effective rate of evolution of SARS-CoV-2 compared to flu. Once you account for that and it's actually four times faster than the flu, it's very fast. Now what sort of happened is that we developed this first generation of vaccines without really taking this into account.
(00:49:11):
And of course this was a picture that, I mean it was relatively predictable from the beginning, but obviously the picture became more clear over time. And the sub gap measure we have is that mRNA vaccines are also quick to update. Our update rate is still too slow. We are probably one major variant behind usually with the updated vaccines. The other direction, which has not really been taken yet, and we have done quite a bit of work on this, is actually trying to do a more highly multiplexed mRNA vaccine. So there's no reason why the mRNA vaccine has to deliver only one version of the spike protein or of the RPD and indeed booster. The motor Pfizer boosters last year were bi valent, so they had the original Wuhan spike and then the omicron spike. The number two could be 20 or 40, especially with better mRNA.
Immad Akhund (00:49:58):
Could you even guess what feature variants might be?
Hannu Rajaniemi (00:50:01):
Machine learning is being used to try to predict the regions where immune escapee might happen, but actually, I mean what's sort of showing up in preclinical experiments around these kinds of highly multiplex vaccines, which are also done sort of in nonna fashion, although I think that'll be non-scalable in practice, but if you show the immune system enough versions of the spike protein, it actually generates more broadly in neutralizing antibodies.
Immad Akhund (00:50:24):
It's just like destroy everything that looks anything like this?
Hannu Rajaniemi (00:50:27):
Find all the common patterns in what you see and then go after that. We have the technology to solve this, actually, we just don't seem to have the will. There is sort of a government attempt project called Project NextGen that is trying to be sort of a warp speed sequel.
Immad Akhund (00:50:42):
When you say the will is lacking, you just mean it's gone back to the default state, which takes 10 to 15 years to go through clinical trials and et cetera?
Hannu Rajaniemi (00:50:51):
Not necessarily. It is definitely dead zone in terms of commercial investment.
Immad Akhund (00:50:55):
Not to be anti-capitalist, but the current state of the vaccine is great for capitalism. It's like every six months you need a new one. Capitalism doesn't necessarily optimize towards like has one vaccine never come back.
Hannu Rajaniemi (00:51:08):
That's right. And the revenue numbers are really insane. Pfizer and biotech I think have now taken more of the market share, but even Moderna is still making many billions a year from the covid boosters. So we do have the technology to make the last Covid vaccine. Just the question of whether the funding ecosystem supports it, maybe the project next gen is the best bet, but they also seem to be spreading their betts across many different approaches. Maybe too broadly actually in my view, but at least some attempts are being made. But I think it is unfortunate that this is where we've landed because there are issues like long covid that will end up having I think a big, big impact on both individual happiness, productivity and long-term chronic health conditions. That feels like a bit of a ticking time bomb that we might only see the realize the real impact of few years down the line.
Immad Akhund (00:51:57):
We've ended up on a relatively non-deadly vaccine variant. I think if it mutated to something more deadly, we might regret not investing in a more generalized kind of solution.
Hannu Rajaniemi (00:52:08):
Most of the population has had vaccines or exposure that help a bit. It is still quite likely that we will see big evolutionary jumps like we did with omicron and the deadlines is complicated. It's a function of our immune response as well as the viral evolution. But I mean we might well see an omicron event where you have a very sudden complete breakthrough through both vaccine and natural immune responses and then everyone gets it and even if the actual death rate or hospitalization rate is not that high, that it'll still have a massive impact on the healthcare system like Omicron.
Immad Akhund (00:52:43):
Do you have a take on whether, I guess sar, COVID two was bio-engineered and or maybe more broadly where is kind of bio-engineered weapons heading based on where technology is today?
Hannu Rajaniemi (00:52:58):
It seems hard to parse the SARS-CoV-2 origin story at this point. There's just too much fog around what happened. It's certainly possible that there was some kind of lab escape. I mean it's striking that the one Institute of Virology is in one and was engaged in coronavirus research and lab escapes do happen and there have been historical examples of those lab escapes causing outbreaks like seventies flu outbreak from a sort of Soviet lab escape. So I dunno, I think probably specifically in the case of SARS-CoV-2, it's like 50 50, hard to say. Certainly not excluded by anything we know or strongly, strongly excluded. In terms of bio weapons, I think that is a tricky question. We are certainly all the sort of exponential things about DNA synthesis and everything around that are definitely enabling capabilities with small scale bad actors that are quite worrying. There are some infra hazards here that I don't even want to say in a podcast on what kinds of things are possible. But yeah, I think we will see engineered pathogens for sure.
Immad Akhund (00:53:59):
And one thing I've read is that it just doesn't take that many people to do it, right? This isn't necessarily with nuclear weapons. You need thousands of people involved, but this is a smallish set of people that can make it happen.
Hannu Rajaniemi (00:54:10):
No, exactly. It is like a decently equipped molecular biology lab, 20 people and then or less, actually not even that many probably. But I mean you need some baseline level of expertise. One concern I think people have expressed with some of the large language models now like Jet GBT, is that they actually might to some extent democratize the expertise you need to generate some of the ideas. I think the actual practical lab work is still challenging enough that there's a nontrivial barrier, but it could be individual terrorist organizations, it could be small, bad actor rogue states. Those kinds of groups are definitely now capable of generating quite worrisome pathogens. And I think one lesson also from Covid I think is that the pathogen doesn't necessarily even have to be lethal to disrupt. I think it can also be an economic weapon. Disrupt supply chains impact people's productivity and decision making because of course the challenge is that when you deploy a bio weapon, your own population typically is also vulnerable.
(00:55:07):
I think theoretically, eventually we might be able to get into more targeted bio weapons, but that's still quite technically challenging. But if you were an actor like Russia who just wants to create chaos and disruption, then you don't necessarily have to release a lethal thing and maybe you can surreptitiously vaccinate your own population in advance. Also, bio whipps might be the nukes together with ai. They might be the nukes of the 21st century and obviously very different fundamentally in nature. So I did recently write a novel draft where the scenario is that, not necessarily that I believe that this will a hundred percent happen, but a possible scenario for the next couple of decades is that we see increasingly fee frequent waves of both zoonotic pandemics and then human made engineered pathogens of increasingly bizarre variety. When it gets to the point where to actually live a normal life, the countermeasure has to be an mRNA wearable.
(00:56:03):
So you actually have a microfluidic chip that you wear that is able to get just digital information, synthesize an mRNA vaccine in C two, and then update your immune system in real time. So it's these continuous software patches for your immune system that are going to get rolled out continuously. That sort of becomes also a useful delivery method for other therapeutics and vaccines. But if we want to really build the global immune system, whether it's sort of that form factor, but I think we do need to figure out how to get much better at defense. People often talk about how bio weapons are asymmetric attacker has an advantage. That is true to some extent, but we also have a defender advantage in terms of our immune systems. Our immune systems are pretty good at dealing with new pathogens if enhanced with things like mRNA vaccines. So we just need to get much better at deployment and development part of it.
Rajat Suri (00:56:53):
What other applications are there for mRNA that do you think is how is it going to change society 20 years from now apart from the ones that you're working on, the cancer, the cultured meat, what else do you think is going to change with mRNA?
Hannu Rajaniemi (00:57:06):
So useful framework that Nicolai and I have converged on is to really, let's think about this as a kind of Carlson curve or an exponential curve, and I'm actually calling it Shanko curve after Nikolai because he came up with this. What is the quantity that will grow exponentially? Let's say it's sort the amount of mRNA or the number of base pairs of genetic information that we can safely deliver into the body. So the size of the X of genome if you like, in terms of base pairs. So let's imagine that's going up exponentially. So right now we are at the one kilobase scale of Exogen roughly. We can take one protein, we can take one thing like a spike protein and we can turn that into mRNA and put it in, or we can replace one enzyme or something like that, one missing thing or put in one new thing.
(00:57:51):
Things will look very interesting if we get to one megabase. So I think in 10 years, and this is not necessarily global deployment, but what is possible and what is in the clinic. So 1 million base pairs equivalent of genetic information delivered as mRNA. That might look like a pan virus vaccine. For example, one shot immunity against all respiratory viruses or common colds, everything. It might look like a preventative cancer vaccine. So obviously something we are working on directly, but something that is broad enough to cover all likely cancer mutations. So that's kind of the mega megabase scale. 20 years after that, I think we get to gigabase scale. That might look like taking all the receptors in someone's immune cells and delivering those as mRNA. So a synthetic, a complete immune profile of someone that then gets you to dealing with autoimmune diseases that there's sort of healthy aging phenotypes.
Immad Akhund (00:58:47):
Wait, so could you take a mature adult immune system and inject it into a baby so they could be immune to every single thing that adult is?
Hannu Rajaniemi (00:58:56):
Yes. And not necessarily a specific mature adult, but what is the optimal immune profile?
Immad Akhund (00:59:01):
For a mature adult.
Hannu Rajaniemi (00:59:04):
Like an AI synthesized version of what the perfect human immune system should look like. Wow. So that's the kind of scale where I think we can get to in 20 years, 30 years, we can get to one tera base, 100 billion, billion pay pairs of genetic information. Then I think we might move out of the realm of mRNA. But to give you an indication of what that scale means, that's like our entire microbiome. Human genome is three giga basises. The human microbiome is actually a hundred giga basises. So our microbiome has more genetic information than the mammalian part of us. So at the tera base scale, we can get to a fully synthetic microbiome. So we could take the perfect microbiome and duplicate that. Now we have perfect metabolism or we can metabolize things that we can't currently metabolize. If you want to eat cellulose and digest, that becomes impossible.
Immad Akhund (00:59:52):
We're going to be eating grass and trees. Yeah, yeah. This is the idea to your next book here, Hannu. Can we get wings now? I want wings.
Hannu Rajaniemi (01:00:02):
If you really want wings, like a functional wings that are controlled by your nervous system.
Immad Akhund (01:00:06):
Yeah, I want to fly. I don't want them to just look good.
Hannu Rajaniemi (01:00:08):
The easier way to do that rather than just actually grow them. Because I think the issue, there's also you would have to completely remake your skeleton, et cetera to be able to biomechanically support that. But I think that the easier way to do that would be to have synthetic prosthetic wings or perhaps lab grown wings and then we ize your body to it. I think one, actually a huge application area we didn't talk about yet for mRNA vaccines is sort of the opposite of generating an immune response against something. It's actually marking something as self. So that's a very rapidly developing area of using mRNA, anti-vaccine essentially to tune down your immune responses for autoimmune diseases. But a very similar application is telling your body not to reject something foreign. So that would then enable you to engraft biomechanical wings or a prosthetic or like a pig organ, which is something that people are obviously also actively working on now. But I think the root to wings is through cyborg wings where there's an mRNA layer to help you not reject the engraftment of those wings.
Immad Akhund (01:01:09):
I want to go back to your 10, 20, 30 year thing. So what's like the 50 year thing if we just skip to the end?
Hannu Rajaniemi (01:01:15):
After the synthetic microbiome? So that was the base scale, so like 20 40, 20, 50. Then we get to PETA base scale. So 10 to the 15 basis of genetic information. We now know that in the brain there are between 3000 and 5,000 different cell types. There was a new sort of major study that just came out. So the brain actually has quite a lot of diversity. Each neuron firing. We often think of the brain as just the connectome, like this neurons connected to each other. But each neuron is a very complex machine and as we have now just learned, there's like three to 5,000 different types of cells in the brain. So the brain state or something like a memory is probably defined not just the connectome, not the connect, just the connections between neurons, but also the state of the individual neurons and the distribution of different cell types in the brain. So at database scale we can clone that so we can then actually clone someone else's brain state at a molecular level. In your brain, maybe you can actually literally download memories or really fully feel what it feels like to be someone else or certainly do very radical radical ai, brain computer interfaces.
Immad Akhund (01:02:21):
Is memory actually stored as a series of cell shapes and proteins?
Hannu Rajaniemi (01:02:26):
There are some really striking observations that are sort of coming out about how the brain stores information. One thing that really blew my mind recently is that in a way viruses are sort of the natural exogen. A lot of our genes are actually repurposed ancient viruses that have come from elsewhere. A really major one in the brain is called arc one. It's basically a virus that has become part of our own genome. It's still able to make viral particles in the brain that package mRNA from neurons and transmit that mRNA to other neurons, and we don't really know what it does. But if it's knocked out in mice, those mice can't form new memories.
Immad Akhund (01:03:06):
Oh, interesting. So maybe memory is mRNA?
Hannu Rajaniemi (01:03:09):
Maybe it is. I mean, or certainly there are more complex biological mechanisms than just like the connections between neurons that are reinforced. But my point is that if we imagine the amount of genetic material that we would need to deliver to duplicate these patterns or the amount of genetic information, then I think PETA base scale I think gets us to actually duplicating all those patterns, patterns in the brain.
Immad Akhund (01:03:29):
What comes after peta pico is that what after peta exo
Hannu Rajaniemi (01:03:34):
Exer, depending on how you count the total genomic information on earth is about an exa base. This becomes a little harder to visualize, but imagine this is now like a post-human form of homo-exer basis where each individual has access to an entire planetary ecosystems worth of genetic information at will. That's when you get your wings, you get multiple bodies, you become an ecosystem essentially that can morph and adapt to different environments and yeah, maybe exist on the same kinds of timescales that planetary ecosystems do.
Immad Akhund (01:04:05):
It feels hard to imagine that's 50 years away.
Hannu Rajaniemi (01:04:07):
That is not necessarily 50 years away, certainly not the practical applications of it, but this is kind of the thing about exponentials. They hit you surprisingly quickly. My argument is that the mRNA is an exponential technology. It is driven by exactly the same forces that are driving other exponential technologies. It is directly driven by Moore's law. DNA synthesis uses the same lithography and microarray printing technologies. DNA sequencing is an exponential technology for the same reasons and the amount of genetic data available to us that can drive things like protein design is growing exponentially and then we add the AI layer to that for the design. I would be very surprised if the change curve of the size of our external genome is not an exponential. So at some point it'll blow up and the timescales, whether it's a few decades or for even a century off, doesn't matter that much in the same way that is the case with AI.
Rajat Suri (01:04:56):
It sounds like you think it'll also impact longevity quite a bit.
Hannu Rajaniemi (01:04:58):
Oh, absolutely. They are. Probably the most promising approach to radical longevity extension or health extension is this idea of cellular rejuvenation or reprogramming. There are a number of companies working on that. Companies like Altus Labs, like a 3 billion Bezos backed new efforts. There's many others new that's Blake Byers and Brian Armstrong kind of an earlier stage company. But the common to these efforts is the idea that if you use these same factors that we use to turn our normal cells into stem cell-like states, but transiently for a little while, that can actually rejuvenate our cells and mRNAs absolutely the perfect vehicle to accomplish that. So longevity is definitely in the cards.
Rajat Suri (01:05:42):
Amazing. Well thank you so much.
Immad Akhund (01:05:44):
We could talk forever, Hannu, but this is super, super interesting as it is. Really appreciate you taking the time and walking us through not just what your company's doing, but a real science lesson in mRNA and the future. Well, Hannu is an amazing book writer. He wrote the Quantum Thief and the Fractal Prints, which I didn't even realize there. He wrote them until this conversation. But they're both great books and he's got four out there with one coming up.
Rajat Suri (01:06:10):
I can't wait to read them. Yeah, I can tell why you're a great sci-fi writer. I mean, you have all these great ideas. You're able to connect the technology to the applications really well and understand how the technology is going to evolve to change the applications. So that's really exciting.
Hannu Rajaniemi (01:06:23):
No, thank you. I appreciate that. One important thing about writing and reading science fiction is that it's sort of easy to think about the impact of technology and sort of abstract context, but if you actually sit down and try to imagine a world where people are using this technology every day and actually touching it and feeling it and interacting with it, you sort of vividly bring that scenario to life in your mind, which I think best science fiction does really well. Then you usually notice things about it that you would not otherwise. You actually understand the human dimension is much better.
Immad Akhund (01:06:54):
Yeah, makes sense. That's why you write great books and make good companies. Alright, we better wrap up. Thanks Hannu.
Hannu Rajaniemi (01:07:02):
Thank you so much.
mRNA and the Future of Biotechnology with Hannu Rajaniemi, Founder and CEO of Helix Nanotechnologies