 Awesome, so I think we'll go ahead and get started and we'll have everyone who's trickling in join us, but can everyone see my, well, I don't actually know if I shared my screen. And everyone see my screen. I don't know there it is. Okay, okay. All right, awesome. Well, thank you for joining us today everyone my name is Trisha to whole ski I'm an associate program officer with the board on life sciences and I'm working on this. This project that we're holding an information gathering session on the topic of biotechnology startups focused on RNA sequencing and modifications. Let's start out with just a few quick housekeeping items before I pass it over to Nick Adams who's going to be moderating our session today. So just an overview of our activity under which this information gathering session is being organized so this is a meeting that's an information gathering session for a consensus study that is focused on sequencing and mapping of any modifications and the study itself is funded the study and the meeting are funded by Warren Albert Foundation and the National Institutes of Health. I'd like to acknowledge our workshop planning committee and our consensus study committee, and also our staff team who has helped us to put together and plan this information gathering session and who are working on the consensus study as a whole. So just a few housekeeping items before we get started on you can feel free to make comments and ask questions by using the chat, or you can use your raise hand function and zoom and we'll call on you. If you have any technical questions or issues you can contact our program assistant name who who's in the room and also you can find his email here. Any comments and ideas made during the meeting should be attributed to the individual speakers and not their organization unless otherwise stated. Any thoughts shared during this meeting should not be interpreted as the opinion of the National Academies of Sciences, Engineering and Medicine, or the committee conducting the study. The meeting will be recorded and is being recorded for use by the committee, and then also finally harassment and bullying will not be tolerated so please be respectful of all fellow participants and speakers during the meeting. And then with that, I'm pleased to pass it over to our moderator Nick Adams who is also a member of our consensus study committee, Nick. So, so I'm Nick Adams I'm a systems engineer at thermal Fisher and a member of this committee. So I'm going to go over a little bit of background to provide context for this information gathering session. This is the committee's statement of task which has been assigned to this committee. If you've checked out the website maybe you've seen this. Just a little bit of context so we're tasked with conducting a consensus study on the needs and the emerging field of epi transcriptomics. Specifically as you can see here it's we're tasked with developing a roadmap for achieving direct sequencing of RNA and its modifications and we're doing this by examining the bullet points listed here. If you go to the next slide the three that were focused on for this information gathering session with those that are invited are related to the scientific needs, the current methodologies and their limitations, and then potential new technologies. And so, before I share these meeting goals maybe I'll give a little bit of context so this committee has been mindful of some of the lessons learned from the human genome project which we see as sort of a reference for us in putting this consensus study together. The human genome project also came on the heels of a national academies consensus study. So we've been looking at that and one of the lessons that we've learned is maybe that the value of industry and in particular maybe startup companies was initially overlooked and became a strong, strong player in the human genome project and we perceive that startup companies will also be a major, major players in epi transcriptomics and sequencing and mapping RNA modifications. And so, our hope is to engage in integrate biotech, the biotech industry large companies and small companies, small companies into this effort which is in part the basis for this information gathering session. And so, here we're bringing together experts in from biotech startups to discuss state of the art for direct sequencing of RNA and their modifications and related therapeutic applications so the goals of this meeting these two bullet points here to understand the state of the art technology for direct sequencing of RNA modifications and then highlight ongoing and planned research development activities by your companies and understand the drivers for innovation. With respect to what we're doing with the putting together this consensus study report we're going to be interested in and we'd love to hear. Just about your company's technologies, current and future projected technologies but also any insights on what drives innovation in your companies insights related to orchestrating concerted efforts between industry and biotech and government, if you have any experiences with that, then also identifying, you know, major technology gaps or challenges that might be required to achieve your goals. So that I think we can go straight into this, the presentations. The five of you will be given presentations like Trisha said Steve Brooke it will be joining later. We'll start with Paul mola. So I'll give just a really brief background Paul while you're pulling up your slides but pause the presidency EO of Roswell biotechnologies. He's previously a leader at applied biosystems and life technologies and launched the ion torrent which is the first semiconductor based sequencer. And he continues to scale biology with semiconductor chips. So with that go ahead, Paul. And Paul I don't know if you've started speaking but I haven't heard anything from you. You may be on mute. They don't sorry about that. Okay, we can hear you now. No, I appreciate the opportunity and about the effort here. So a big part of my career really has been figuring out how to integrate biology onto chips. As Nick indicated, I have done this through various other various other opportunities specifically at life tech where we actually launched the first semiconductor sequencing technology. And then I was also involved with some of the work that was done at junior where they were interrogating single molecules are using nanopause and that's what really led me then to myself and my co founder to start Roswell. The idea is that we're trying to figure out how to go to all electronic sensing or all electronic bio detection. And that allows us then to be to be on chip and have a system that is very scalable, but also very robust in terms of its applications. So just some of the fundamental design principles we followed is, you know, one is we need a system that can help us achieve really low cost which will help with the sort of mapping that, you know, this this you're talking about here, for example, having some $100 chips and very low cost transcript terms and very low cost sort of methylation profiling as well. To do this we, like I said we need to be on chip in a way that is scalable which means the sensor itself needs to be a nano sensor so unlike some of the other technologies like iron torrent or junior, where they deployed the sensing element itself was a micro, but then it was deployed on a nano technology which is the underlying CMOS so there was a bit of an incompatibility in terms of how that scales. And so we specifically needed it to be nano. We wanted to achieve single molecule resolution as well. And this will become important as you think about having the resolution then to read methylation, which we believe is actually you get better resolution if you can read single molecules versus ensembles. And then the platform we created is really also very programmable and that means that it is also capable of multiomic sensing as well. DNA proteins, etc. So with those design principles and also based on some of the previous experience that we've had with bringing other chip technologies to market. So we found it then Roswell and I'll introduce the sense itself in the next slide. So, let's consider any two molecules that interact, you know in biology this will be antibody antigen this will be, you know, DNA strand with its complement, any two molecules that interact in nature. But what we do is we take any of the two, and we wire that onto this little circuit here which is essentially a current meter and wire one of the pro one of the molecules as a probe onto the circuit. So these really these are you know metal electrodes at the end here and then we use a in our case in this example, we actually use a peptide or very specific sequence that conducts, you know conducts electricity, and we have metal binding domains at the end and so we can program it and actually make it part of the circuit by you know conjugating the probe onto onto it. And so this is, this is again the probe conjugated onto a peptide bridge, and then we click that into a current meter. Now, the way this works is, you know, consider this particular unbound states to have a specific, you know, resistance. And what happens is that when it binds its complement, you can actually change that resistance. And so the readout is essentially current versus time. And we can actually see that dynamic interaction of you know in one state when it's not bound it has, you know you have a given current level. Then when it is bound, you see the current level jump up in this case and so you see these on off states, which are pretty much then the dynamic interaction of that probe with its target. And so you have the real time dynamic view of the interactions. And again this probe target could be pretty much any biomolecule you choose to have. So once we achieve this and actually saw that yes we're getting the readout we, we, you know, we desired, we then took this circuitry. And then we deploy that on a standard semiconductor chip, which is shown on the right hand side here this is actually an SCM image of our gen three chip this is it has 16,000 sensors on it. Now, each pixel, which is sort of the blue dot here is represented and arrayed multiple times in this case 16,000 across across this chip. And so, because now it's on chip. And also we've achieved what we've been trying to achieve which is molecular sort of molecular size, so to speak, in this case, the gap sizes around 20 nanometers so it's very amenable because it is nano scale to deploying on a nano technology which means it's actually very scalable and so the scaling limit then really opens up and you can have very high high throughput chips, so to speak. So, so with that, we then just to give a sense of how, how do you program this little probe into the circuit. The way we do it is, is we have this, again, let's call it a molecular wire, and we have this probe conjugated onto it. Then what we do is we can either wait for this to self assemble where the probe, you know, will move by diffusion, and then we find the electrodes and click in place, but that will take time and it's not efficient and so what we do is we actually use electrical forces. We create a dialogue dielectric field, in this case show here, and we can actually polarize the molecular wire, and that draws in and it clicks in place and we complete the circuit. There's one below here, this is an open circuit so there's no wire and we do one, two, three, four attempts of active bridging these are very fast, you know, cycles of active bridging. And once the wire in starts in place, the bridges inside and you see a jump in conductivity, and that particular electrode is actually bridged and we observe that in this case using. We have, we call it dumbbell bridge but these are just sort of gold, gold beads that are attached to the wire and we're not to visualize them, and you can see it's clicked in place. So we use this. Can I ask a quick question is this, are there is there an aqueous solvent in this. Yes. Okay, thank you. Yes. Yes, that is. So once this is programmed then you're ready to to run your assay and and and get going. Here's an example of how we we programmed the bridge so in this case, because we can access each row. One at a time individually and polarize it one at a time and bridge one at one rather time. I'm showing an example here where wrote two to seven, we choose not to break so there's no wire here. So we serially then row a to 13 we we wire in a bare molecular wire so in this case it's just a wire without the probe. And then in this case for 14 to 49. We have a wire with a specific in this case it was an antibody. And then we sort of repeat just for the sake of having controls to show you what the data looks like a bare wire and no wire. And these are actual traces. And the way we round the experiments is fast. Once we've wired the chip or program the chip, we just flowing buffer and you can see it's just there's no signal at all it's just flat. And then we start to flow in the target, and we do this with I trade the target just to absorb it. And you, you, you do it in increasing concentrations from low concentration of high concentration. And you see that the chip, or the, the, the rose or the pixels that have a specific antibody bound to them when I introduce increasing increasing concentration of its, of its active target. You can actually get these sort of signals and you can then use that and carry that to analysis which I'll show in the next few slides. So this is how we program the chip it's a very fast process. This can be done at your site, or it can be done at the factory if there's specific sort of probe molecules you like that is it can be done either way, but it's very fast and efficient process. Here's an example where we have a probe as a rather DNA oligo as a probe. And in this case we're really trying to assess what it how does it interact with its complement target. And so this is the unbound state. This is the bound state. And what we do is we're showing here again just a very short about six seconds of data, where we show this dynamic interaction of the on off state as the probe or oligo binds its complement. And the way this data looks is that the bound state usually shows up in this elevated sort of current versus this is the unbound state which is a noise bound and you can see very two distinct states that are showing up here. Now, I don't have it in this slide, but if we do introduce a mutation in this, the signal, it does change the kinetics and we can actually observe it so it's very, very sensitive even at a single base pair mismatch level, and we can see various sort of interactions, resulting in different kinetics and different profiles. This is all label free. There's no dies. You have single molecule sensitivity. Like I say this is a very short sort of example of a few seconds that it's very rich in information. And again, it's, it's very scalable as you need to. Yes. So you say it's very sensitive single molecule. Is it the chemist don't like this is it sub stoichiometric to say if you had one molecule in the hundred that had the modification of a certain location would that be something you'd be able to pick out and quantify this technology. And you can, I couldn't speak to one buses a hundred but if you did have them at a given ratio you should be able to. We are using this for example to do I'll start binding where you're looking at different states of the same target interacting in different ways and we can actually pick those up. Thank you. Sure. So Paul I just want to check in. We're going to have about two or three more minutes of slides and then we'll leave about five minutes for questions. So maybe I and I don't know how many slides you have left but if maybe questions can be added to the chat and then Paul could answer them throughout the rest of the meeting. Thanks. So I'll be fast but again this is just showing you how we look at the data. We derive the full kinetics, you know, the, the time bound time waiting, the profiles of the picks, we can then use that then to, to get your single your sort of classical measures and derive the care of which we use a lot of analysis derived from that. Here just showing that you do have concentration, sort of how concentration shows up where the wait time is longer if you have a lower concentration versus it's shorter when you have a high concentration and that's how we do some of the analysis. I'm going to jump to the slide that specifically speaks to methylation and in this case what we do is we actually wire into the circuit DNA polymerase enzyme, and the DNA polymerase is conjugated to the molecular wire again the peptide bridge. And what we do then is we introduce then a DNA strand and then introduce the basis and and then the DNA starts to add one base at a time, and that results in incorporation spikes that we then read out. And this is essentially. Let me show this slide. This just shows the physical path of the electrons and how we think the signal is being derived, but it's a full sort of the conformational change as the DNA polymerase incorporates a base. All that conformational change and the, and the, the, the charge effect of the base itself is what gives us the resolution to to call out the sequence. So, here you go. In this case we run a homopolymer of just a's and a homopolymer of C's this is one strand of DNA. As you can see, we can absolutely resolve homopolymers very well. The a spikes have a very different phenotype from the C spikes which is shown here, and we can actually discriminate in this case just show that with greater than 90% accuracy of a's versus C's in this case. Now, when we introduce sort of a methylated sequence in this case, we had a template that had sort of a methylated region buses and methylated, and we can actually be able then to resolve and at this point, you know, where we are in the state of the development. You know, 99% or 100% accuracy but this is still an evolving application development, but we do see that you do get very different profile incorporation profiles of native buses methylated basis, so to speak. So, you know, there's a paper we've published it's a PNAS paper. It's our seminal paper where we've shown a little bit of this data, but really excited that we'll be able to to actually deploy and bring this technology to market in a way that is quite transformational. So with that I'll just pause and answer any questions. That's great. Thank you very much, Paul. All right, so any, any questions for Paul. Okay, so I maybe I'll start so this is so this is really interesting and compelling data that you have on. It's focused on DNA methylation do you. And I know, you know, startups you have to be focused, and, but do you do you consider opportunities for looking at RNA and and also RNA modifications. And have you thought of strategies and how you would do that. Well, I, as an organization, our strategy has really been that we, we want to enter the market in a way that is quite differentiated. We see some of the work that is being done by the Grails of the world and and and and we believe that the solutions they have are really not scalable sort of for patient population use from a cost perspective. And so yeah, so it is a very interesting area for us to be able to provide solutions around methylation but also in a way that you can read these sequences in from a sort of multiomic signature that includes methylation as well and it's an area we are really interested in because we feel that we are quite differentiated to make a big dent and difference in this from an application perspective. Yeah. Sirath has a question. Yeah, so Paul, great talk. So, I'm curious if the association between signal and the nucleotides I may have missed it. Is this a one to one or two to one meaning at any given point, the signal is corresponding to how many basis. So the signal. Each, each spike is an in is an incorporation spike for one base. And so each basis it's getting incorporated relates to the full sort of transformation of the, the polymerase sort of opening up grabbing the enzyme and incorporating it and clicking it in place. It's a one spike. And we adjust the frequency of the read to to sort of match that incorporation spike and you can adjust the read frequency so that it's really higher resolution or lower resolution. And then what we've also done in some cases is actually modify the base itself with a charge so that it's a charge level base, and then you can actually be able to even either enhance or quench the signal depending on what the application is. Brenda. Yes, thank you very much I liked it, the chip format and I in your chip format is, you know, multi of you have different molecules you have. I'm wondering, are they do you need multiple channels for each sample or do you envision these are multiple samples on the chip. I envision that these are going to be single sample one time use and toss it out. Right so it's sort of a disposable chip. But you bring about good question and the way we think about it versus multiplex sort of samples or patients on it to increase the, the resolution, we can actually have different polymerases in targeting the same sample in different ways so that you can get a richer data set so different polymerases may give different signatures for methylation. And so then you will pull all that data to get even more higher accuracy, higher accuracy, sort of methylation profiles. Thank you. Yeah, there's ways you can introduce lanes as well if you wanted to do multiple samples as well. Great. Okay, so we have one more question and we're about out of time but Byron, go for it. Yes. So my question was, I really, I really like the approach everybody like the technology. It was interesting that you're talking about differentiating differentiating from other technologies as it looked like there were some overlaps between the nanopore situation which is built specifically for DNA and they haven't optimized an RNA pool and also depicts us who do this sort of magnetic RNA characterization effort. So I was wondering how what do you see is the main differentiating factor between you and those two technologies. Well, I think in the case of nanopores, right, they rely on you need to floss the DNA for the four multiple times and you don't have necessarily direct single base resolution, you have to use algorithms to extrapolate because you have three to five bases in the pool at a time. In our case it's truly direct read of one base at a time. And then from a scaling perspective, the nanopore resides in the lipid bilayer so we feel it's a bit of a macro on chip, which is what Genia was doing, versus this is truly at a molecular scale. So from a scaling perspective, we believe this will scale better. We believe that we have a better potential for higher accuracy because it's a direct read. And then the way we can array also very different polymerases, it's more natural for a polymerase to be in targeting sort of a DNA versus it's a natural for the DNA to be flossing through a pool. It didn't evolve to do that in nature, versus DNA polymerase does that in nature. So we just feel it's a different better approach. Thanks, Paul. Alright, so I think we'll transition to the next speaker now if you have other questions put them in the chat I know Paul there's already one in there if you want to access the chat maybe you could respond to that. The next the next speaker we have is Martin Huber is the founder and CEO of quanta for Right. I also have a question for Paula but I put it in the chat for later just Let me roll in this one. Is this the full screen now. That looks perfect. Okay, perfect. So, I'll jump right into it Nick, if that's okay. Yeah, go for it. All right. Well, thank you very much for for the invite and my name is Martin Huber I'm the founder of quanta for and that we have developed a nanopore based DNA and protein sequencing systems are Disclaimer maybe upfront. The one thing that we haven't done yet is RNA sequencing and I know this meeting is about RNA sequencing and modifications of RNA detecting modifications. But I talked to Nick and Trisha prior to this meeting and we found it may be interesting to this committee to learn about our technology. And we already had some some discussions about RNA sequencing actually after that initial call and be, you know, be able to do that with our system we just haven't focused on it. So, as I mentioned, we have developed an optical nanopore based approach to read DNA and more recently, we realized that we can also use this to analyze proteins sequence of proteins on a single molecule level. Maybe quick background as Paul prior to starting quanta for it was a giant torrent. I was much earlier than all of us the third employee at I am torrent here on the west coast. And after the acquisition through life technologies, I realized that there's a lot that can be improved on the sequencing. At that point DNA sequencing with respect to sample prep read length cost, etc. That kind of motivated me to start quanta for to build a multiomics platform capable of reading reading DNA and eventually proteins. I think I can skip through this slide here. Everybody knows the numbers and as you can see what okay. The RNA is missing here. That's because I can be focused on DNA, then realize we can do proteins or the protea forms. But I promise we'll, we'll look at RNA the system is capable of doing it. So, I want to start off with talking a little bit about the over DNA sequencing approach how it works, and then move on to the to the protein fingerprinting. So, hang on a second. So our, our sample prep or library prep is a little bit simpler than you know the classical sequencing by synthesis approaches in a way that we don't need to amplify the DNA. Our sample prep is a primer extension reaction where you have the DNA strand, the template strand that's shown here in gray, a primer that binds to the DNA strand shown here in green. And that can be directly on the sequence or to an adapter that's ligated to the DNA strand. And then we have a polymerase that copies the template strand and it incorporates fluorescent labeled nucleotides. The polymerase is an evolved polymerase. It's not a natural polymerase in order to be able to incorporate these modified nucleotides. The polymerase copies the DNA. At the end falls off. We do a purification step to remove an unincorporated nucleotides polymerase primer and so on. And then as a last step, we attach the double-stranded DNA via biotin that's attached to the five prime of the primer to a nanoparticle that's stripped of it and modified. And that's it. So the whole process is relatively quick to extend. I mean, it's like a single cycle PCR, if you will. The extension is we usually do it for five to ten minutes just to make sure that everything goes to completion, the binding to the nanoparticles is also the same range of few minutes and then the DNA is ready for sequencing. We then add the nanoparticle with the attached labeled double-stranded DNA to our nanopore chip. And you can see here a cross-section of a single nanopore. There is obviously a multitude of nanopores in our chip. These are solid-state nanopores, so no proteins or bilayers involved. And you can see the cross-section here of a single nanopore. There is a little bit of simplified depiction here. We have a silicon nitride membrane and metal layer and we apply an electric field across the membrane. So the next nanopores it's here. There's one over here, obviously. And then by applying this electric field, we attract the nanoparticle with the attached DNA and pull the DNA into the nanopore. The nanoparticle itself is bigger than the solid-state nanopore. And with that, the translocation stops when the nanoparticle gets plugged into the pore. And then in the transbuffer, which is this part here, we have an enzyme, an exonuclease that starts digesting the labeled DNA strand. The labeled nucleotides that are released are diffusing then towards a laser excitation zone. So we have a laser that's illuminating the bottom part of the chip continuously. The metal layer prevents the laser from penetrating through the nanopores. So we don't have any photobleaching happening inside the nanopore. And the released nucleotides are then detected by a CMOS camera. That's part of the sequencing platform. Let me just go back here. So again, the electric field attracts the nanoparticle with the DNA. The DNA gets pulled into the nanopore, stops the translocation, and then the exonuclease is digesting the labeled DNA. And the labeled nucleotides are then diffusing out of the nanopore into this laser excitation zone and generate fluorescent signal and photobleach very quickly in that process. The monster exonuclease has digested the DNA. The signal stops, right? There's no more signal generated. We then turn off the electric field and the nanoparticle diffuses away immediately and then we turn on the electric field again and we reload the nanopores. And that cycle then is repeated. And that's how we generate multiple reads per nanopore. So Legal told me I can't show recent data. So this is some very early data that we generated where we used two different colors. But the signal looks pretty much the same. It's a little bit more complex from different templates. So this is an artificial template that we used. So the template is the top strand. This is where the primer binds to the template and then copies the DNA where we have these. With the red label, the ATP and the green label, the UTP, the exonuclease will bind to this side of the template. So it's attached to the nanoparticle over here via the 5 prime biotin and then the exonuclease will digest DNA in this direction releasing first green signal and then the red signal which you can see here. So each of these spikes is a single nucleotide. So similar to what Paul mentioned before, we also or unlike, I should say Oxford Nanopore or any other electrical nanopore sequencing platform. We have single base resolution since the exonuclease releases one base at a time. And it's a base per second on average. You can see that here it's roughly 100 nucleotides in length or 80 nucleotides in length. This template so we digest one base at a time. And the signal to noise is fairly, fairly good. So they're very high accuracy, you know, for single molecules it's about 90% of this at this point. So this is how the DNA sequencing works. So then very quickly before a run out of time, we didn't realize that we can use the very same technology to identify, not sequence the proteins. We call it protein fingerprinting by labeling certain amino acids to the fluorescent diet. This is how we do it. We take a protein of interest or a mix of protein for that matter. We denature the proteins completely using SDS heat reducing agent like DDT making it a completely linear chain of amino acids. And then we recently labeled certain amino acids. For instance, lysines, cysteines, methionines, tyrosines are relatively easy targets for very specific attachment of fluorophores to those amino acids or the side groups of those amino acids. And then we also attach a biotin specifically to the N-terminus and we use that to attach the labeled denatured proteins to the very same nanoparticles that we use for DNA sequencing. And then, excuse me, we use the very same nanoportship with the same metal layers, the same sequencer using a laser illumination. We then pull the nanoparticles with the labeled proteins attached into the nanoportship. Instead of the exonuclease, we use an exopeptidase in the transchamber. And then, as soon as the protein gets inserted into the nanoport, the exopeptidase starts to chest in the protein, releasing the amino acids and the labeled amino acids we detect similar to the DNA sequencing as they exit the nanoports. So, right now we have three amino acids labeled shown here. We don't detect, again, the unlabeled amino acids shown here in black, but we get a barcode or a fingerprint of a certain protein. And I may use that to compare it to a reference sequence in a database that allows us to identify which protein we just, which protein we digested. And since it's single molecule, it's obviously a quantitative approach. We can tell how many or how much of a protein is present in a sample using this approach. So we did a little bit of bioinformatics behind the whole protein detection system just to see whether our platform would be able to, or how much of the whole proteome it would be able to detect in a single run that's shown here. We took the average human proteome and with 100,000 nanopores, which we can put on one chip and illuminate simultaneously by depleting the most abundant proteins shown here up to 100 of the most abundant proteins. And running, doing 50 million reads, we can get up to somewhere between 80 to 90% coverage of the human proteome. And that allows us, you know, with a fairly high dynamic range to also detect the low abundant proteins in a sample. Obviously, more of those if we remove the most abundant proteins through filtration in the beginning. Martin, just a time, I think we're almost there within about a minute. Okay, I'm almost almost done. So this is just, we started building our system that the cartridge holds that nanopore chip. We use standard nano fabrication processes to, you know, wafer scale processes to generate our nanopore chip. We fill the software that runs all the components of the instrument, but also does the analysis. We use machine learning to identify the signals. And as the amino acids and nucleotides are coming out of the nanopore and it will, you know, offer some sample preparation reagents. I think I talked about all the other things. And with that, I can answer any questions if there are any. Great, thank you Martin. Yeah, any questions for Martin. So when you're detecting individual amino acids or nucleotides, I'm presuming you're using software to build the full length molecules. So do you anticipate facing challenges with isoforms or homomorphic repeats associated proteins or single molecule RNAs? Homopolymers, sorry, did I understand? So the homopolymers are no problem, because the, again, we just did DNA sequencing so I can't really talk to RNA sequencing, but you could imagine if you had the polymerase that can copy on RNA, so on RT, we could potentially direct the sequence RNA. And the exonuclease in the actual sequencing reaction digest one base at a time, right, so any homopolymer. So we never observe any issues with a homopolymer or an increased, you know, error rate based on a homopolymer. So quick follow up Martin and you're using solid state nanopores, right? Correct. So are silicon based nanopores or what are these? Yeah, there is there is multiple layers. So yes, it's a silicon, silicon nitride. Then we have a certain other layers for, you know, adhesion and all kinds of, you know, from a nano fabrication perspective, and then we have a thick metal layer that we that basically prevents the laser light of penetrating through. Think of it as an upside down zero mode waveguide, if you will, the laser light cannot penetrate through the nanopore, which is, you know, crucial because it does this prevent any photo bleaching while the DNA or the protein sits inside the nanopore. So sorry, sorry, what's the shelf life and reusability of these pores, because the issue with biological nanopores is, once you are exposing them to current, they get degraded, right 24 to 48 hours or thereafter. So I'm curious about the shelf life and reusability. So shelf life we've used chips that are years old, and they perform just fine. And we have done runs 48 hours with, you know, no increase in current so that the pores don't, you know, increase in size or anything. So we don't see much of it. So we haven't gone beyond that, but you know, that's pretty 48 hour run this is probably won't go longer than that anyway. Thank you. Okay, Brenda. And so I'm going to ask you to extrapolate to RNA, which is maybe not fair but you can pass if you want, but I guess I'm thinking of methods that for RNA that involve exonucleases to cleave off each base, and it seems to me that, you know, especially if you have modifications, you have a chemo methyl, you're going to have a problem sort of with getting the exonuclease to work the same way on every modification and I just wondered if you have any thoughts on that. Yeah, you're probably right. That would be my first thought that, you know, with modifications on on RNA that an exonuclease would have probably, you know, would react differently. Let's put it this way. Again, we haven't done any any any work with RNA. So that's why I'm really just speculating here. I would, I would expect it to behave somewhat differently. However, I mean, so the way we would, let me just very, very quickly jump back here to the, to the one slide where, you know, we did a. You know, the way we would envision RNAs, instead of DNA, right, using having here an RNA with potentially some modifications to the RNA, and then we would still need to somehow, you know, incorporate those labelled nucleotides, right. So we would basically, you know, convert the RNA into a cDNA or a hybrid RNA cDNA molecule where the cDNA is modified. And, you know, for that we would need a specialized, you know, polymerase, right, an RT polymerase reverse transcriptase. And I'm not even sure if that polymerase could actually recognize specific RNA modifications and transfer that into the cDNA. So that might be another, you know, challenge for RNA sequencing. But then the exonuclease at that point with that chest, right, the labelled cDNA and so. Right, yep. Yeah, that's a good point. But you would have to have some way of the new cDNA strand being able to read out a different modification. Yes. Thank you. That was really helpful. Yeah, that's perfect. We were right at 20 minutes. So thank you, Martin. We're going to turn it over to Sammy Jeffrey. He's a professor at Weill Cornell Medical College and founder of Gotham Therapeutics, and an advisor at 858 Therapeutics, which has recently acquired Gotham. Right. Thank you very much. So, you know, and I just wanted to mention that, you know, when I'm speaking today, I'm speaking as a, for my university position, I'm not speaking as a representative of any of the companies that, you know, I've been involved in. But I'm going to tell you today about our work on, sorry, the wrong timer. Just my experience working with companies that want to drug the epitrinscriptome, why they want to do it, what the challenges are that we're facing in terms of charging the epitrinscriptome. And, you know, I think why were people interested in this? You know, we were first approached in 2016. You know, after we did the original map of M6A at relatively low resolution 2012 and then single nucleotide maps in 2015. Many companies approached me and I think the column, and in 2016 I said, you know, why? All we know is maps, you don't know what it's doing, but their idea was, this is the new epigenetics and we want to get in early. So the column group and verse and ventures, I think I was involved with both of them, starting with Accent and then starting with Gotham Therapeutics. I left the column group to join verse and to do Gotham Therapeutics, which was then acquired by 858 to brought in the RNA portfolio. But it was a very lucrative, lots of money was raised in the series A without that much of an understanding about what these modifications are actually doing. But the idea was at least the pitches that these companies were giving is that there is a whole universe of modifications out there. M5C, pseudioridine, M1A, 2-prime omethyls, all these other things in mRNAs, and each one just like every histone modification could have a unique function. And a lot of these were also in big, huge, high-profile papers. Our original M6A map was in cell, but all these other things were also in big journals as well. But we now know that most of these things were just artifacts and at least 99% of the map sites for almost all of these other modifications, especially M1A, 2-prime omethyl, M7G, and maybe AC4C. AC4C was an artifact, but nonetheless at the time created a very different perspective of what you could target, which was every different modification would cause every different function. But we now know that a lot of them are not real. And why aren't they real? A lot of them were very simple errors. I mean, I think the three most well-known problems are that 2-prime omethyl, which was, you know, we and others showed that they were mapping their reverse transcriptase primer. The M1A modification, which was in nature, we and the Schwartz group showed that the antibody was not specific. The M7G, Schwartz has argued, this is a great review on the problem with the mapping of these modifications and the errors that they've all been about. But they showed that it's the alumina sequencing errors at the ends of reads. The problem is that people did not use biochemical validation, especially the method that Taupan innovated called scarlet, which we use all the time in other labs they use, but none of these papers and none of the reviewers or the journals really required it. But, you know, the thing, the ultimate thing is, regardless of whether, and I should say we've also done other methods to see if there's other modifications. We've done these global analysis of huge amounts of sequencing reads to see if there are recurrent mismatch errors at specific locations in the genome relative to the DNA, because that can indicate the presence of a mutation. A lot of mutations will, a lot of modifications will cause mutations or reverse transcription. There's a lot of sequencing errors, but if a specific site shows errors over and over again across different replicates, and we screened 3 million nucleotides. And we only find five sites which are showing mismatch errors during reverse transcription, one of those turns out to be M1A. So there's very, very few modifications outside of M6A. M6A does not introduce a sequencing error, but the thing is, even though there's been a lot of evidence showing that these modifications don't exist, that they've been large, that they're largely artifacts, the VC people are still so excited by it. And so it's, even now, even after the initial fund raise, I still get accosted by venture capital people about targeting these different modifications. And even this, this was also like an example of something that was written in Nature Genetics, a call for these modifications. But yes, one person said to me who's in the field said, you know, why are, this is like saying we want to study unicorns and leprechauns. These things don't exist in RNAs, these modifications. There's no real need to really map them if they don't exist or if they exist in very, very, very, very, very low frequencies or abundances. So the diverse mRNA epitranscriptome doesn't exist. It's really M6A and a few others. I think we can, I basically put the modifications in three different buckets. So one bucket is the real and the abundant modifications. And then there's M6A, the cap methyl, the two primal metals at the first two positions in RNA and that N6 methyl at the first position. So these are the abundant ones that were all discovered in the 1970s. But then there's some rare ones that exist, like pseudoyuridine, maybe 100, M5C, maybe 200, M1A1 or 2. And then there's others that are not validated and are probably artifacts or are also super rare. So the thing is, there isn't really a diverse epitranscriptome, it's really M6A. And there's all these cap methyls and we mapped the M6AM in 2015 and we just published the cap methylation map a few months ago. So I think those things are it. There aren't really anything else to map, at least in a reasonable number. They're just super rare examples. So that's why we have a focus on M6A. And all of these companies, all the companies I think are pretty much focusing on M6A is because it's the only one that's really there. And M6A also has a clear effect, mRNA degradation, specific RNAs are also modified with M6A, which are cell fate and differentiation genes. And so, you know, if you want to modulate cell fate, which is important in cancer, that's a great thing to target because you can target a coordinated set of genes that are all related in a single pathway. And even some patient data, there's some mutations in some of the pathway proteins as well as overexpression of metal III and certain leukemias. So there's a reason to believe that M6A would be very good. On the other hand, pseudo urethane and M5C, which do exist in very low numbers, even after all these years, there's no clear and consistent effect on what they do. There's no clear reader. And what I think you're going to find is that every pseudo urethane that you might find has its own unique story, like this pseudo urethane may affect binding of protein X and this one might affect translation and this one might affect stability. Each one will be different, but there's no general principle like you have with M6A. Now it might be different in viruses. There's some interesting evidence for modifications in viruses that are robust, but in mRNAs, it's not really a clear thing. So that's why I don't think, you know, I or anybody recommends going after these modifications. M6AM and cap 2 methylation, these cap modifications, you know, even though I love these because we map them. It's not really clear what the disease links are either and it's still early days, but I'm not really sure if there's a valid reason to go after those modifications as well. So it's just M6A. And then the question is the companies are all talking about what do we go after? And the thing is that there were so many readers, writers and erasers. The idea is that you might be able to select any of these and select and micro surgically dissect the M6A pathway. But a lot of this has turned out not to be valid either. One of the ideas that they're, you know, in terms, let me just go back here. What I'm showing you on the readers is that these YTHDF proteins are all different readers and the idea sort of was indicated at the bottom that each of these YTHDF proteins targets a different M6A and that they do different things. I'm showing you some clip data where people are looking at the binding sites in the top row YTHDF1, the bottom site YTHDF2. And you know, you can see, look how different they are. You could go after, we can make drugs that target each one of these. But these all have turned out to be mostly artifacts as well. I think a group out at Duke showed that the YTH clip data that had been used, the reads were only six nucleotides in length, and so you can't map them in a single way. So we redid the clip analysis and we were trying to see, based on the peak heights, we took every dot on the right is M6A site and we took the peak height for YTHDF1 and the peak height for YTHDF2. And we said, are there any M6A sites that are preferentially binding to one or binding to the other? The answer is no. They all bind equally and we showed that all of these sites bind equally. You know, this paper I think was probably the big paper that really demonstrated this idea that the different readers do different things and that the YTHDF1, which is what this paper was about, does translation, but there was simple errors in these experiments. And so I'm just showing you the ribosome profiling, which this paper was based on. They were doing knockdown of YTHDF1 and you can see the number of reads for every gene. And the reads should not change except for YTHDF1, which is what happened. But in their second replicate, it looks normal except the YTHDF1 was actually massively overexpressed. So they didn't do a knockdown, they did an overexpression and this sort of created an artifact in terms of our understanding of YTHDF1 because of the way this was done. So we redid all of this work and the conclusion was in our paper in 2020 that all the YTHDF proteins are the same. They bind all the same mRNAs. They do the same function and there's no argument to make isoform specific inhibitors. So that I think this paper, our paper really ended this idea that we should drug them separately because they all are basically doing the same thing. And this work has been now replicated by Jacob Hannah's group, other groups. So, you know, this is kind of well accepted. There might be some differences, but they all do degradation. So anyway, so we now have a very, very different view, at least the drug industry has a very different view of the M6A pathway. One writer, which is turns out to be the Metal 3, Metal 14 heterodimer, the claims that Metal 14 was a separate enzyme was due to contamination in the way they did their preparation. One major reader, the YTHDF proteins that are paralogs are almost identical. They do almost the same thing. A lot of the other proteins, you know, they've now reported crystal structures. They don't actually bind M6A. And one eraser, FTO had been argued to be the M6A eraser. We showed that in fact it doesn't even act on M6A in cells. It acts on SNR and A, but there's still some debate about this. I think different people have different views. But I think we're going to come to a resolution on this very soon. Alkbh5 is a true M6A eraser and it's in test use. It's not really clear if it has roles outside of test use. But there isn't really the universe of so many targets to go after. The M6A pathway turned out to be very, very, very narrow. And so it's very different from how when the initial fundraising happened to as science developed, we now just know if you want to target M6A, you can target the readers. But then I think, you know, there's such a big issue of like we want to get rid of M6A and MIC, or you want to get M6A in P53, but we now know core ideas that were present early on are not really correct. So first, all M6A sites in the transcript don't essentially every single one is mediated by this metal three enzyme. So you can't target one set of M6A sites by targeting one pathway or another. Any inhibitor targets all of them. And all of the sites do the same thing. So, you know, you can't target any individual YTHDF protein because they all function to mediate the same property which is degrading all RNAs that have M6A, at least M6A's and high abundance. And then the other thing is that there's no demethylation by FTO, and I think the Schwartz group did FTO knockouts, Shugo did FTO knockouts. And there's a lot of, and there's, and we analyze this ourselves, but the methods I'm giving you are quantitative methods. No, no real change in M6A sites. There's a paper that argues that maybe it's in retro transposons, not in regular mRNAs, but at least in mRNAs, there doesn't seem to be any effective demethylation. No tissue specific M6A. And so this, and you know, a really great experiment in some of these papers has been to compare different tissues, very different tissues, differentiated and undifferentiated stem cells. And they look at the M6A stoichiometry at every site and it's the same. And so this is why I tell people, and you know, we invented M6A mapping, but I tell them, don't waste your time mapping M6A. It's a universal code. And you can look at any one data set and it will tell you if your M6A is modified into the exact percentage. And I think the real thing, you know, we argued this in molecular cell in 2022, but now some papers in molecular cell science have now more or less proven this, that the M6A is hardwired by the genomic architecture. And that's why our large internal exons are what recruits the methyl transferase. And it's because the exon junk, at least according to some of these papers, the exon is not our work. This is the work of other groups. The exon junction complex hides a lot of the RNA from methylation, and only the accessible sites can be methylated. And that's going to be true in a liver or a T cell or brain because they all have the same genomic architecture, which is why M6A sites are universal and they're changing and pretty much identical in pretty much every tissue because they all have the same genomic architecture. So where did this idea that M6A is dynamic come from? Why do we even have this idea? Well, it's because of the way that people have incorrectly done their M6A calling. So if the top row in purple is a myripsych map where you get, myripsych is the method for mapping M6A sites and you get peaks on the genome, you can say, oh, and you have a threshold and a dotted line. You can say the first two peaks are called M6A sites, but not the third one. And then in the cancer, you might say, oh, the second one is not called, but the first and third one are. And you'll call the second one as a unique site that's lost in cancer. And the third one is a unique site that's gained in cancer, but that is absolutely the wrong way to do bioinformatic analysis. You shouldn't be saying a site's there or not there. You should basically compare them. So these Venn diagrams that people have used have really misled the people into believing that there's cancer-specific epitranscriptome when there's absolutely not. I talked about this and I think this paper really talks about how the peak maps are so noisy and you can't be doing this type of analysis. And pretty much all the work in this field, and I'm sorry to say it, all the work in the field has been using the incorrect way of comparing the sites. But let me just tell you, all of these problems have now been solved. We've worked by two groups or two sets of groups, all in nature biotechnology late last year. We now have quantitative mapping of M6A. It's these methods basically deaminate A into in a scene, but they don't deaminate M6A. And so you can just see any adenosines that remain in your sequencing reads or M6A. And this is just an example where adenosines are in orange. And the only place where you see any adenosines in your reads was at the M6A site. And the difference between the two methods is blurry. Does it do a chemical reaction? ETMC uses an enzyme to deaminate all the adenosines. But the key thing I just want to say, they've done their quantitative analysis and they compare, for instance, stem cells with fibroblasts, completely different cell types. M6A stoichiometry identical at every single site. So are basically identical at 99% of the sites I should say. And the other sites are probably just because of low read coverage. So this just shows that we now can get exact stoichiometry since this is the method. So when I deal with a biotech, you know, the key conclusions are from these works is that any metal-green inhibitor will target all of them. There's no disease-specific M6A FB transcriptome. It's the same as a normal FB transcriptome. There's no evidence for one RNA, like MIC being 10 times more methylated than any other cell type. Or at least the evidence is very, very flawed. And so the focus of targeting M6A should be when you want to target the entire M6A cohort. So the thing is, I think one of the things that we deal with is the FDA wants a biomarker when we make the metal-green inhibitors. People have all agreed, okay, I'll wrap up and I think in one or two slides. They want biomarkers and there are enzymes that cut M6A modified transcripts and you can do QPTR. This is using the MAZF enzyme. But the key thing that I emphasize, people are not very happy with what I say because they know that FDA won't like it. Any M6A site can be measured. You can measure M6A levels and act in. And if they go down when you treat your animal with the inhibitor, then you've inhibited the enzyme. All M6A sites are identical. You can test any one of them, but they do prefer to use disease-relevant biomarkers. And, you know, we were brought on to help people do mapping in these companies, but there's no need to do mapping with the metal-green inhibitor. All sites are affected. The prospects, I think the major gap that we have is how do you, and it's not really technology, it's science. How do you know which patient is going to respond to metal-green inhibitor, which one won't? For mRNA, I think the future involves the real abundant modifications, but M6A is the only one that's disease-linked. The real untapped areas, tRNA modifications, a lot more dynamics in the modifications. There's a lot of potential links to disease, especially Richard, I've already shown some of this. RRNA modifications could potentially be interesting in regulating specific targets of translation, and they may be disease-relevant, and SNO RNAs guide those modifications. But we really need evidence that these modifications are driving the disease. All right, so just these are my acknowledgements, and I can take any questions. Great. Thank you, Sammy. I'm sorry, I lost my camera, but I'm still here. I'm sure there's going to be lots of questions, but maybe, are you able to tell us a little bit about Gotham? I know that at the beginning you said you didn't represent Gotham, but given your, you know, insights, is there anything you can say about how what the work they're doing fits into what you just talked about? Yeah, so I can talk about what they publicly disclose, and the key thing is, you know, again, they were trying to figure out a good target, and the answer is metal-3. That's what they decided to go after. And the YTHDF proteins are something to consider, but they don't have really good drug-able pockets. So even if you could get an inhibitor that inhibited all the YTHDF proteins, they just don't have the right features for drugability. So metal-3, which uses acetylenosomatine, it has a pocket right there, and there's a long history of drugging these enzymes. And so the main focus of Gotham was to target metal-3. And we were acquired by 858. We have a broader RNA portfolio, and so we're targeting other RNA biology. I don't want to say RNA modifications because I want to be general because I can't reveal any of the programs that they have other than the ones that they publicly announced. But they're going after RNA biology, and it fit in very nicely. It was very synergistic with a lot of the other expertise that they have. So they're making inhibitors that we're going to be in clinic, I think, at the end of this year, or at least we'll have the IND submit at the end of the year or early 2023, 2024, I mean. And we're closely following the great work from storm therapeutics, which is already described in metal-3 inhibitor, and they're already in phase one trials. Yeah. Okay, great. Thank you for answering that. And then so, so Byron, I'm going to prioritize Brenda, not anything against you. Your questions are great. But just to get the committee question in and then we'll see if we have time for yours. I'm not sure I like that, but anyway. So, Sammy, in the methods that have been used to map M6A, I guess I thought they were mostly ensemble, you know, that you're not getting end to end for single messenger RNAs. And I, you know, when you say they're all the same, maybe that's right. But if it wasn't all end to end, you would not be getting diversity in where the modifications are. And then finally, I will just say, of course, in a scene in messenger RNA has been linked to disease. Yeah. Yeah, sorry, you know, you know how it is in the epi transcriptomics field. We all kind of ignore in a scene in 8i editing, yet that was actually first. You know what I mean, that was actually well before I had this stuff. I do know what you mean. Yeah. So let me just say Brenda is absolutely right. So if you see using some of these methods that is 50% M6A, it could be that at five sites, let's say there's five M6A sites and they're all 50% stoichiometry. It could be that among 100 transcripts at every site, it could be randomly yes, no, yes, no 50% or could be half the transcripts have all have M6A at all the sites and half have M6A at none of the sites. The ETAM seek people who use an enzyme and there's no RNA fragmentation are doing long reach sequencing and we're doing it as well. So I think that will soon be resolved where you'll be able to look at individual RNAs and you'll be able to just see if you get a transcription of RNAs which are completely M6A free and then sometimes you get transcription with M6A full, or if it's all stochastic and somewhat random to achieve the final concentration. But it's basically, I mean, wait a year, you'll get all that information. I mean, these methods are revolutionary. I wish we developed it we actually were thinking about it a lot, but, but then there was a surge of the aminase enzymes for CRISPR base editing and these people use that and so we I should have restarted that program in our lab. But that that's a great, great, great method, and the chemical methods pretty good to it fragments the RNAs you get small reads but if you want long reads, the ETAM seek is the way to do it. But you know the key thing is no more mysteries. Well, I, yeah, I will let other people ask questions I, I got your point and your, you know, you put it forth very well I guess I'm not sure everybody would agree that it's all the same but Well, let me just say this, I think the problem that we have is that so many people have written papers arguing for diversity and dynamics that it's, it's, it's hard to say that, you know, all of this stuff is wrong and my colleague Chris Mason who wrote this paper we were part of it but you know he basically listed every single paper and showed how they were wrong and I told Chris, you're never going to get this into nature by technology and it kept getting rejected because I think it was being reviewed by people whose work he was showing was incorrect but I think that the reason Brenda why I think it's going to be wrong at all is because the data is so un, is so unambiguous when you use these methods you can't fight about it anymore. There's no more like, Well, I saw a peak change or I didn't see a peak change, you get an absolute percentage and so I think this will allow us to answer all of those questions. And I think the data will be hard to deny these these papers they're not mine but they're really really good. Um, I have a lot of questions that they'll take way too long and I'm also really hoping that we can give Byron a chance to ask his question. But so I'm thinking in terms of you know the tasks of the committee and thinking about, you know, sequencing of RNA modifications and obviously your, your opinion is clear of which ones to go after. So thinking about even just m6a, you know, you're making the argument that everything's kind of the same, yet there is evidence that m6a can have specific functions on the processing of transcripts that are different from RNA to RNA, and obviously I mean m6a can have different physiological effects in different cell types etc etc. I wonder if maybe we can just hear your thoughts on, you know, the advantages of an effort like this ie, you know, direct sequencing of abundant chemical modifications like like m6a you know what's the value if, if we're to think that everything's the same and it doesn't matter we could just sequence one RNAs m6a. Well, I always tell people do not waste your money and sequence m6a it is a waste of time, because every m6a site that we have found that we mapped in our single nucleotide resolution map comes up and every other person's data set the only difference is that different people use different thresholds and then they call a site more or less and it's related to RNA abundance and other types of technical things. These assays over here will solve that right. Again, they've already provided the ETAM-Seq had many different mouse tissues, the glory has different sites and human and mouse. I mean I don't really think there's any need to do it except to resolve the residual controversies. I think Brenda's point about looking at them in single molecules is the only remaining question that we have. I think what you're going to find is that and this is what we find that m6a levels globally go up or globally go down a few percent, maybe 10%, 20% depending if the degradation pathway for m6a RNAs is more or less active, but that's it. But it's not like there's different patterns of writing. Also, I think in terms of m6a is having different functions, there's a, you know, Neil Brockdorf had a great paper on m6a and splicing and he used a chemical inhibitor before they were available and showed that it was only like three or four mRNAs but they had dramatic splicing changes. But in general m6a's are going to be linked to one thing, instability. There are effects on translation but they're always indirect, right, because m6a can influence the expression of translational regulators and other factors and then you can get changes in translation. I don't think, especially now that we've got chemical inhibitors and this is what Neil Brockdorf, you know, Neil Brockdorf's paper was like, look, when you knock at metal three, the cells are differentiating and then when you look at splicing changes, it could be because of a different cell type, not because you inhibited m6a. I think now that we've got the chemical inhibitors and we can acutely measure what's happening, like is there a different splicing effect? Is there a difference in some other aspect of biology? I'm going to see there isn't that much, there's going to be a few unique circumstances but it's broadly one function, I think it's going to be degradation. And I mean, and that's what you see globally, right, global effect of degradation. I mean, and let me just say, Kate, there won't, I'm not saying that there won't be one off papers like in this mRNA, the m6a's position in a way that blocks protein X or does something. But as in terms of a general function, that's what m6a does. It's like microRNAs, you know, some microRNAs actually turn on mRNA stability and some microRNAs compete with proteins and you can get a paper and look at your cell showing this microRNA does this thing on this mRNA, but the broad function of microRNAs is to mediate degradation. All right, so I'm going to cut in. We have about 41 minutes left for two more speakers so I think that works out well for us we're going to cut off the questions but I did see that Byron asked a question in the chat. And you can answer that while he's given his talk, I'm going to turn it over to Byron, he's the director of RNA analytics at Storm Therapeutics. Hi Nick. Yeah, and I don't have any slides so sorry. This is, I'm just talking, I invite questions as I go because there's nothing for you to interrupt, there's just my flow of consciousness as it were, but yeah, we'll try and stick to time. So Byron, that sounds great. Let me just interject, if you can mute your mic, I think there's some background, not you Byron, but others, if you can remember to mute your mic, it might help us hear what Byron has to say. It might be some people having a lovely time near me, but yeah, I'll do my best. So I'll just quickly refer to Storm Therapeutics and a few numbers with 35 people in Cambridge UK founded seven years ago in 2016 published about 25 papers. We've raised $85 million of venture capital and we have one candidate drug in the clinic, which is STC 15 on metal three. So yeah, that's that's that really can't hear you Byron. You can't hear me. We can't hear you. Oh, it came through. It came through okay for me. Is anybody else having a problem? I can hear it. Okay, sorry Brenda, I guess, I guess you're out. Oh, shout. Can other people hear Byron? I can hear, but there's a background conversation. Okay, I'll just be away from them. Let's see if that works. Okay, let's get this based away from them. So yeah, yeah, so in general, we've got a metal three in here, but that's that's not really what this is about. I could talk a little bit about my experience again. This is not really what it's about. It's very interesting to follow Sammy and a lot of assertions and things that I really enjoy about M6A and metal three in general. I would say, there's M6A. There's a huge amount of work that's been focused in on that for technical reasons and I think quite a lot of what somebody says is, is correct essentially I don't agree with all of it I'm nervous of simple solutions to complicated things I don't think that every M6A is the same. Yeah, maybe we can get into that in questions and debate, but I would say what is probably as important is to try and characterize to work out what you're talking about with the M6A. There's sites which are consistent. There's sites which are stochastic, as is, this is my belief. This comes from the various studies we've done of all the different kinds of sequencing technologies. We've got eight people in the group used to sequence by mass spec, we use nuclear side mass spec internally as well. We did some nanopore stuff. It did all sorts of various sequencing things for us I think I would agree that the main place of effort to go for what what has like the highest return on interest for understanding the diversity of the RNA mods is probably in some other forms of RNA than the messenger RNA. So I think for the messenger RNA mods, whether you agree or disagree on the spread or the occupancy or the reality of them. I would say that working out what happens downstream of them is almost the more compelling thing that will take a really long time to work out that's just time and it's a lot of effort. I'm not sure it's something that can be short circuited that easily. I think that's just a very long time and much research from a lot of people. I think understanding the diversity of the modification profile on ribosomes and on tRNA and how that differs across diverse biological settings might be something where sequencing the modifications may be more immediately help they might provide answers rather than just more questions. So that's that is a guess my view of it. I'm happy to take any sort of debates and questions matters we go. But I would say, as well as the aspect the RNA sequencing the cell biology this various stuff that I believe that lead at storm. I also direct our Adel one discovery program which I work with x Alexis, there are lovely collaborators in San Francisco. So what I would say is that quite a lot of the learnings that we've, we've seen from from the inner scene is that we can we can treat some of these things some of the consistent in the scene sites the ones that everyone's interested in. Treating them as simple self non self markers goes quite a long way towards what you end up then understanding about what happens as consequences of messing around with them. The similarities between M6A and in the scene is itself non self markers and treated on somberly system. The sequencing them help you to understand that we need to treat them all on mass as some kind of mass pharmacodynamic fire marker. I mean that that's most of what we do. But like what I would say is, I'm going to circle back on the M6A thinks this is fun to so so much opinion from Sammy so so much to chat about with it that's great. I guess to give my personal takes and feelings on it. We did nanopore sequencing we did loads of it did it industrially because like we can. And you get some sites that pop up all the time everywhere every, you know, every, every experiment that you look at, but not always and with the sort of algorithms we're using at the time we're using safe but boring ones that were working with the biophysics not doing anything particularly clever or neural networking. So we're working with like 400 sites that we could always always always see and these were antibody free that limited by sensitivity because you need at least 100 reads going through each biological nanopore together. So sometimes the chips die and it's really boring. But in general, we ended up with the idea that we could, we could do clever things with data and we could get 100,000 sites back with extremely low frequency and we could do these less clever things and we could get these 400 sites we could track through some time and whatever. And I think that they behave that they exist for different purposes, but we can't really match the assertions we can't mutate them all. And I think it's something for mcpc in particular. Very few people go after the mods for mutation studies afterwards because it's fearsome it's threatening it's horrible. A paper that really affected me is written by someone online I think I said Kate's written loads of papers that I like, but I was really struck by the single cell Darcy paper. The idea that I find it very appealing the idea that metal three with the short degenerate motif has no structural components to it I mean like who's ever heard of a motif that isn't made out of canonical nucleotides as that sort of slack in it. So the idea of like the exome junction complex getting in the way. The m6a kind of just been deposited wherever this ends up happens to be this relatively misunderstood writer complex of loads of different bits through it as well. I think the idea that there's a huge amount of stochastic noise that we couldn't see with an anniversary concert because that's not within the remits of the technique. That m6a still matters is something which we've struggled with and we think about quite a lot. The point of where I'm sort of marching with this I guess is, if you were to try and sequence all of the transcript, you're going to try and find modifications that matter. Do you need to do sick cell m6a seek to try and find the level of information that's in there, or is that kind of impossible are you constrained to those constant m6a sites that that are few and for probably reasons and transcript tonic architecture are always in the same place. Or for m6a in particular is the challenge more about working out what happens ensemble to removing. So I think decoding the specific m6a locations is something that you can rationalize is really hard. I think other techniques that validate the single cell that seek if they exist would be required for that. I don't know what they are. I think that the place where you might get more answers to reverberate my own point is to say, if we're looking at specialized ribosomes or something like this, some of them reabana work which I also find really exciting. Can you say why some ribosomes are different to others. Can you look at the code on usage stuff comes out of that. Can you sort of explain why something that's thought to be homogenous, but actually isn't. There's more value there immediately but whether those techniques are as a meanable to some of some of what we were hearing about earlier on. I'm not sure. That's probably talked about 40% of stuff I was talking about half the people can't hear me willingly accept any questions on that at all and maybe I can shout a bit louder. Thank you Byron I think it came through clearly and Brenda does have a question. I realized that I really don't know where m68 is outside of mammals and I wonder if you could remind me if I think I remembered to software but see elegans you know how deep does m68 go in messenger RNAs and is there has anybody look to see if it's the same between mammals and other animals. Wonderful question and I think I'm used to know it's really well about five years ago when at the time when all BCs were saying, are you going to be able to stratify patients by understanding their m68 locations and I still get asked that question and now I just say no, it's not that easy. Sadly, but I think yeah there is an equivalent to metal three in see elegans. I believe there's an equivalent to metal 14 as well as it sort of this obligate heterodyne relationship. I don't know about all the other, your vermin's your W taps your RBM 15s. These other components I don't know how can serve the whole complicated complex is. But I do know that the main place where people have really understood what m68 is doing in any sort of dynamic way I think is yeast really but I think there's been these studies where they go it going up and down at different points of the cell cycle but it's not stuff that I look at that much as a drug discoverer I have to say. Yeah, so that's a good point looking for the, where the enzymes are the readers writers is a good way to look and, and maybe there's not as much sequencing information for other other organisms but I will, I will let one go on. No, so the question and point that I wanted to make is that clearly and somebody did a good job of listing the modifications that so far have been ascribed to mRNA. And by hook or by crew coming some of them may or may not be artifactual. But the fact that you limit the description to those modifications or it speaks volumes, but also the fact that maybe we just don't know how to see others. So we are just limited what what technology offers now, but maybe by the time some is about to retire and say wow, this crazy new technology from Byron or whoever they get is now shows that indeed that are like 50 of them that we just simply failed to the day. And this is the story because I mean, I'm going to bring down also a little bit about why I'm saying this mean when modifications already described. It took 30 years to to to discover 30 and then sat there. Actually from one to 30 to 30 years with nothing happening. And even though the technology was so bad they still got to 30. And then 1995 of course is where I'm rigorous and calls the golden age of modification. So he jumped to 160 or 120. And now we're counting 185. And it is not just here and there was some other names, non-coordinator names and so on and so forth. So the point being that I, I, although I do agree that M6 says it means the mapping methods for M6 are greatly improved. They do not address it as a single molecule that is clear and end to end, which actually could provide a source of diversity and variability. And secondly, we don't know that we are there yet because how can we know when we don't have the methodology as the point that I wanted to make. And it's a lovely point and I get to steal the responses it's in my slots that's great. I think I was really affected by your talk when we talked about our RNA and RNA modifications can be thought of as a response to the metabolic milieu around them. I quite like that. I think for a lot of the cellular studies were received from these moths. It's a shame they're all been grown in exactly the same media in this really consistent way and I think some of the, I've seen some work from Tom Suzuki which changes the carbonate and also the tRNA modifications and stuff that you can rationalize it makes sense but no one necessarily changes their work close to take account of that. So I think that there, there is, I find it really surprising that most of the modifications appear to be in bacterial species it's not really how I was raised to believe the bacteria are more complicated than the humans at the time because it doesn't really seem to match the sort of increasing complete complexity of genomic pollution. But yeah, I don't know but I think quite a lot of the things that would change the experiments to capture these things aren't things to be routinely do because everyone's still arguing about what M6A does because that's got so much mystery to it as well. Now I can just say something I know this is mostly Byron's, it's Byron's time but one we do, we do know what's in mRNA, you know it's by mass spec and Frank Leico had a great paper where he took mRNA and kept on purifying it after purifying it and found that all these modifications that were there other than M6A would go away and that they were probably contaminating tRNA fragments and other things. Byron there was something else, you should get a new mass spec peak because mass spec tells you the mass and mobility and we know the modification. Now it doesn't mean that there aren't rare modifications, biology is messy, and I think you'll find every single tRNA modification will, enzyme will every once in a while in one plate, one cell on a plate will accidentally hit an RNA and you'll get a little bit. But we know that there's nothing substantive at large amounts based on really really good mass spectrometry other than M6A. I don't think mass spectrometry for mRNA is actually really really good, but I will say the following, but the point that I make is more philosophical that you don't when you don't know where you are, where you are going you know where you're, if you got there yet. So they, for instances, as an example of this, cyclic T6A nobody knew that it existed, because everybody was purifying their tRNAs, tRNAs in this case, in trisbulfur, which essentially destroys it. And it was only when when Susumu and Shimura whispering Tom Suzuki's ear that it may exist if he changed the buffer, it was found. So that speaks also for how we prepare things, how we treat RNA, right. And that's the sort of point that I'm trying to make and not arguing with you and simply saying that the chapter is not closed yet and that's why we are here. Yeah. Yeah, did you, did you want to say anything to wrap up? If there are no other questions we can just wrap this up. Yeah, it's probably a good place to wrap it up. I was probably going to say some things that I can't say so yeah that's fine. I'll wrap it up there. Okay. Well thank you so much Byron we really appreciated your perspectives. So let's go ahead and move on to Steve Bruick I think I saw that he's now on so as Steve as your screen. Just a brief intro Steve is Professor Emeritus at the University of New Mexico, VP and CSO at Harmonica Technologies. Okay, so let me the first preface is I didn't understand half of what was just said because I don't know anything about genetics. My background is optics, metamaterials, semiconductors, lithography, so nano fabrication kinds of things so you'll hear a lot of nano fabrication and no genetics for me. Just real quick Steven so your talk would have fit well with the first couple of talks which were related to these similar things so don't feel like you're too out of place it fits in really well here. Yeah. Okay, so, so a couple of advertisement slides and then I'll get into the meat of it. We, we're using Ramon readout and Ramon I'll talk about that in some detail but Ramon gives you, you know, all the vibrational frequencies of the molecule, so it can tell you all the modifications. We have a potential of very long read performance. We're using surface as Ramon was nanomaniac scale localization. And one more. This is pure vapor where you do not have a box, we have laboratories. But but this is what it's going to look like some. Some of the things we think we can do and I'll just skip over that and get to me. So this was our initial, initial plan that we were now somewhat changing it but but this is still gives you the basics of it so we start out with getting DNA into a nano channel, and then getting it through a porous nano port that slows it down dramatically and then we read it out with a with an enhancement structure at the top. And what you're seeing here that the spectra you're seeing whoops, sorry about that, the spectrum you see over to the, to the right are a CTG and five methyl C in monolayer concentration on enhancement structures and you can see they all have distinct spectrum. So the features of this platform are electrophoretic control of the of the of the translocation, which is by the voltage applied across this whole thing. Nano channels to prepare the single strand of DNA tortures nano porous to slow the DNA enhancement structures which are now metamaterial enhancement structures to provide single base localization sensitivity. And it can be parallelized because we have lots of that you know we can separate now channels by optical resolutions of anatomy of a micron and have thousands of now channels at the same time. Okay, so Ramos scattering the promise of Ramos scattering is that it provides the molecular details, the vibrational signatures are clearly sensitive to molecular bonding. It's a unique identification of all epigenetic variations. The problem is Raman is a weak effect. The bulk spectra which I'll show you in a minute require thousands to millions of molecules and spatial resolution is set by the focal spot size which is much So our approach is to use surface enhance Ramos scattering, which has demonstrated single molecule sensitivities, but usually in non reproducible geometries and was difficult manufacturing manufacturing tricks that are hard to replicate. We believe a manufacturable enhancement structure that provides the enhancement you need spatial localization down to on the order of an animator, and has demonstrated single molecule sensitivities which I'm going to try and prove to you, and can be integrated into So these are bulk Raman spectra of a number of different variants, variations, you see ACT and G at the bottom and then a number of other epigenetic modifications and the important point is they all have unique spec. Okay, so here's let me talk a little bit about our optimization structure. We started out with a gold disk. And we did. And what you're seeing over on the left hands, everywhere on this page or modeling results. And you get very strong enhance books, very strong enhancements. But you learn a couple of things. So if you're doing, if you just put gold disk down on a piece of glass you get a very big enhancement everything is wonderful. But in the real world you need a sticking layer, you need something to hold the gold onto the substrate. And that's what people use and we started out using metals like titanium or nickel, and that kills the enhancement so so you can see that here if you look at this curve right here this is beta beta is the the field at the hot spot divided by the incident field. So you need betas of a few hundred to get to single molecule sensitivity. You've got a you've got a very strong beta, and they tune with the size of these things with the diameter and the height and everything else you want to add. But you can see that you get a strong enhancement if you just put gold down but if you add the sticking layer, your screw. You have nothing, almost nothing. So we learned how to use SAO two is the sticking layer and that that improves life a little bit. Then we looked at, and I'll show you how to get it all back in a second. And that's from this this overhang of the, the pedestal so if instead of. So if you look at these are mim structures and I'll talk about those in a second but but if you look at the bottom there's an essay to sticking layer that's the same size as the pillar. It's overhanging by five nanometers on each side and here so we're hanging by 10 nanometers on each side, and you can see that you get only enhancement back. So, so this is the point is that subtle engineering can can get you a strong enhancement and you have to understand what you're doing in detail to get there. I showed you a mim mim is a metal insulator metal structure and the big advantage of it is that it gives you a broader resonance because you have both an electric dipole resonance associated with the gold disk and a magnetic dipole resonance associated with the loop that's created. And that allows you to tune one of put one of the stokes and one of the palm. And then, then if you go to elliptical structures you can further further concentrate the fields to the lightning rod effects. And, and we should, we'll, we haven't done that yet but we are expecting much higher enhancements. So here experimental verification to that so here's a the first is a disk. These are the simulations these are the experiments, and you can see that there was a disk we get a very we get a marginal signal. We get a bigger signal. And when we add what we call men plus we get a much bigger signal and the men plus is just adding an extra layer to bring the, the, to create the overhang that I mentioned. And if we compare the size of the disk these are again theoretical calculations on the left experiment on the right. And you can see then the, the, well the, the blue curve is is the theory is the beta square beta squared at the pump times beta squared of the stokes that we extract from from these simulations in the experiment of the red dots and you can see reasonably good agreement. Okay, so now, is it enough. I didn't emphasize it but the enhancement is fairly broad in the circumference around the disk, but it's less than a nanometer, both vertically and away from the disk so we in two dimensions we have confinement to, to atomic dimensions. So what what you're seeing here is is monitor is spectra from dilute solution of ANC that that has been been soaked and then rinsed so so we're looking at adsorbed molecules. And in this case it was one sixteenths of a monolayer, and we see every spectra sees ANC and you can see the ANC the this is the a this is a C. And you see them in every spectrum. Then we went to one over 512. And you can see that now, roughly 50% of the of the spectra showed nothing. And, and you're seeing more, more single as more more spectra only show a or only shows the. So we go to further dilution and this is down to one over 1024. What you see is is what looks like single molecule sensitivity. So now 90% of the spectra show nothing. But ANC show up distinctly as separate and ANC shows up in a smaller ratio. All of this is consistent was plus on statistics so we think we're seeing a single molecule. And what we're seeing you on the left hand side is just if we look at a only or see only we see a only or see only. Okay, now we're moving to a more manufacturer platform. What you're seeing here a 40 nanometer channels written by e beam lithography with what we call a pinball structure to unravel the DNA gets it in it goes in through here. Now downstream where we put an enhancement structure. We are now learning how to do the flow control and these structures, we have yet to fire it in anger and to actually see a Raman signal from from the DNA that, but we know we can tell the DNA and hang it here and we're confident we're going to see strong We are a small company. Very small we have nine people. We have four locations Albuquerque in San Diego, we're venture capital funded, and, and that's, that's basically our story. That's great thank you very much Steve. Are there any questions for Steve. I had a question. How, how are you planning to actually do the sequencing I mean so many of us is there a polymerase that incorporates nucleotides or just flowing the DNA past that Brahman specs to your general. We're, we're, we're hoping, I don't know planning is the right word but hoping that we can control the motion with was post electric fields. And as the DNA flows by your right you generate you generate the Raman and we want to ratchet it with with electric fields we know we can stop it with an electric field we can stop it in the nano channel. And we're we're planning to ratchet it but we are clearly going to have problems with all polymers and issues of sequences of stretches of agencies. I don't know if this would be single stranded DNA or can you do this with this right at the end of this single stranded DNA single stranded right. Yeah, yeah. Okay. Cool. Thank you. Yeah, one go ahead. Yeah, I have a question so so in the, when one talks about any modification, you know, that is 185 different. So then, do you then see a case where the limitation of Raymond spectra is that some modifications are going to give you very similar readouts. I won't guarantee that we'll seal 185 but but and actually the problems which I didn't talk about are impurities and other things they get in there and confound your spectra, because we're working close to us close to a limit of singles and right now the spectra take us 10 seconds to take that's way too long to get enough information so we need to, which is why we're trying so hard to improve the sensitivity, the, the enhancement factors. Our goal is to get to 130s of a second and to have the full spectrum but we do have problems, particularly with carbon impurities to get in there in part of the special region and block and when they show up, you can't tell what's real and what's, what's an impurity. Thank you. We don't, we don't have a, I don't have a good answer yet, but, but I think if we can get the full spectra, Rahman has the capability of not not every, every peak will be unique, but the whole spectrum is unique. Thank you. Any other questions for Steve. So Steve, I have a question so I, and I may have missed this in your slides. But maybe you can speak to it. So what, what do you think is sort of the theoretical limit in terms of how many molecules, you need to get sufficient signal, like, is it theoretically possible to get down to a single molecule read and get sufficient signal to determine the sequence. And could you speak on that term for just concentration requirement. Yeah. Well, that's what I was, I guess I didn't get across, but I was hoping to convince you of that is that I can see a single either A or C at this point is all we're looking at in that experiment but but we were down to a single a single molecule. Yeah. And, and so that that speaks to the spatial localization, as well as to the signal strengths. And as a, so this is, I, as, as I understand the field this is the first time anyone has demonstrated non resident Rahman in a, in a manifestable structure at a single molecule level. Yeah, that's, yeah, that's great. I did, I did, I know you were building towards that and then I, I missed it but that's, I think that that's, you know, one of the challenges in this field with RNA is that we don't have, you know, there's no ability to replicate the molecules and, and also replicate the modifications with. And so that's really exciting. And what will the other limitation we're facing at the moment is just getting the getting that that RNA or DNA strand into the, into the right place of course, and we have some strategies for that but we they're yet to be proven. Great. I think we have a little bit of time and there's a question that was added to the, to the chat here I'll just read it from Colleen here. They were asking about how was the data that you showed on the RNA modifications taken. It was a bulk Rahman, or was it single molecule. Well I showed both bulk in single molecule. The, the, the, the spectrum that I showed. Let me go back to it. These are bulk Rahman. But, but, and I didn't even comment on them, but let me find, but, but these are the these, they're not single molecule, but they are on their monolayers on our enhancement structure. And then we looked monolayers, typically we think about 300 300 molecules contributing to those Rahman signatures. And then we see the same Rahman signature, but not was as good as signal the noise of course when we get down to one molecule. Right. Any other questions for Steve. Okay, I think we can wrap it up. Thank you all for joining. I certainly learned a lot, remain optimistic about technology development in this area. And I'll turn it back over to Trisha. Yeah, all I'll say is thank you all thanks to all of our great speakers today I think we had a really interesting and stimulating conversation about where we're going. It's a great place to thank all of our committee members. And if the committee members wouldn't mind just staying on a few extra minutes, that would be really great. So thank you all. Thank you. Bye. Thank you. That's everyone. Yep, we're good. All right, folks. That's exactly what we had intended with this session, but I think it was good information.