 Good morning, everyone. Welcome back to the second day of this workshop. I apologize for a little bit of a late start. We, for those of you who are in person you understand for those online we had some hiccups with traffic and getting food and speakers in on time so we're back on track now. Opening remarks from our other co-chair, Takjip or TJ Ha, and then we're going to launch right into our first session so TJ. Good morning, everyone. I'm Takjip Ha, I go by TJ from Johns Hopkins University. I'm co-chair of this committee together with Brenda Bass from Utah, and I'd like to thank all of the staff members of the National Academies for organizing and supporting this exciting event. Okay, so we had a really wonderful series of sessions yesterday ending in the session titled major concerns and pitfalls in the field and I would say there is about discussion on the scientific need for the sequencing and mapping of R&M modifications. Today we will start with a session focused on more on the societal need for an effort focusing on impact on biotechnologies and disease. In the final session of the day and the morning will be on some brand new technologies that we may not even know can be used for mapping R&A modifications, but we believe will become part of the effort in the future. There may not be ready or mature enough to be applied to problems at the moment, but may become essential. And in between we'll have a breakout group. I believe it will be organized among the in-person participants separately and separated for the virtual participants. There will be another activity a few months from now doing the breakout groups and brainstorming sessions on a larger scale. So for that we are announcing in public ideation challenge. And this is about performing collaborative brainstorming and problem solving sessions among teams of scientists and innovators from different background about how to identify and come up with solutions for challenges in the area of mapping and sequencing R&A modifications. It'll happen in June over a three day period and application for ideation challenge has just opened up this morning as of now. And it'll close in April on April 10. And it may happen fully virtually or it may happen in a hybrid format with some in-person participants as we are running this workshop yesterday and today. And we invite all of you and other innovators to participate and hopefully we will have people from all different sectors and different backgrounds and also from different career stages, including postdocs and graduate students. So, I think it provides several interesting and helpful motivations why you want to do this. You'll get an acknowledgement officially for your participation, especially for the younger participants. It can be a nice addition, one line addition to your CV. It will give you an opportunity to shape the future research in this space and it'll be a great networking opportunities for everyone. And for some of the ideas that are selected for additional elaboration, the teams will be offered some compensation for writing a commission paper that will become part of the report in one format or another. So please apply. We have the URL and code here and this information will be available through email announcements that you will receive and also project website. And especially, I think it'll be really interesting to think about encoding your students and postdocs and in your own labs and other labs in your institution to apply. So I ask Katherine to come to the podium to make a separate announcement. Hi, everyone, I'm Catherine as long I'm an associate program officer here at the National Academies, along with staffing this project. I'm also staff on the Committee on use of race, ethnicity, ancestry and ancestry as population descriptors and genomics research. And so yesterday, our report was released so as you can see the report is titled using population descriptors in genetics and genomics research, a new framework for an evolving field. So as you can see up there, there's two QR codes. The first one you can scan to get to our report includes our report, a quick highlights page, a page with all of our recommendations and then finally an interactive page which includes an interactive decision tree for genetics and genomics researchers. And I'm sure also RNA researchers if you use descriptors to decide which descriptors would be best to use for your particular study. So the second QR code there is for our report release briefing, the webinar that we have on Friday. So that's March 17 from 1130 am to 1230pm Eastern time. So you can scan that QR code to register. Again, please join us on Friday for a briefing from the co-chairs, as well as time for Q&A from the public that will be answered by the co-chairs as well as a few other members of the committee. So I'll pass it back to TJ to kick off our first session for today. Thanks for your attention. Thank you, Catherine. So, okay. So the next session is a session we titled Framing the Day. And our speaker is Dr. Eckhardt Jankowski from Modona. Dr. Jankowski is the vice president of RNA science at Modona. Previously, he was a professor at Case Western University and the director of the center of RNA science and therapeutics. So now we are moving on to the next session, which will be shared by Dr. Susan Bersega from Yale University. She's joining us remotely. Thank you, everyone. Thank you, TJ. I'd like to introduce myself. I'm Susan Bersega. I'm a professor of molecular biophysics and biochemistry at Yale University and the Yale School of Medicine. And I'm here to moderate the session on disease and societal impacts of RNA chemical modifications. So our goals are threefold in this session. And that is to understand the information that is known about RNA chemical modifications and how they're connected to disease. And to explore how patient populations are impacted by expanded research on the understanding of disease, especially as it pertains to RNA modifications. And we're interested in the impacts and diagnostics and treatments as end products of this research. We'll have three different speakers today. And each will give a five to 10 minutes presentation. At the end, we'll have, we'll invite everyone to submit their questions. And for those of you who are online, like I am, in the chat, Jessica has put in the link that you can use to submit your questions. So without further ado, I'd like to introduce Kathy Lu. Kathy is an assistant professor of biochemistry and biophysics at the Proam and School of Medicine at the University of Pennsylvania. Dr. Lu's current research focuses on the co-regulation of modifications across mRNA, tRNA and ribosomal RNA and cancers, and sex chromosome encoded protein homologues in sex by itself cancers. Kathy. Thank you for the introduction. I'm here. Hello everyone in the room and on Zoom. Let's talk about RNA modifications in disease. So like Christine introduced yesterday, both DNA and RNA carry chemical modifications started by the enzymes. But one thing very unique to RNA is we have so many RNA species, and more than 150 types of modifications are a together with a sequence RNA, they can securely diversify gene regulation, because the modifications, they will move. They will. Okay, because they will nearly impact every single step of the RNA processing. Today, my goal is to use this three major RNA species to tell you the modifications function dysregulation to set up the stage for fuel and data to tell you more about the RNA modifications in disease. I like to start with the tRNA because this is the most heavily modified RNA species. As you see here, all the yellow squares are highlighted disease relevant to tRNA modifications. I call them the disease relevant to tRNA modifications is because the enzymes are modified those signs has been implicated in human disease. In general, the modifications stay and the anti codon loop region, they could modify the fidelity of translation speed of translation with the modification on the DRM region of the tRNA molecules they could modulate the conformation stability of the molecules. Let me show you some specific examples. The first example is a elongator complex, which is a six member protein complex that add this carboxy methyl group on your cell to other enzymes that couldn't further modify the structure. This modification occurs at the 34th position of your cell in tRNA the anti codon loop region wobble site. It will enhance translation of codons ending with adenosine. So, to promote codon dependent translation elongator complex is required for normal development. However, it can be hijacked by cancers. For instance, in melanoma elongator complex the elevated level of that has been implicated in patients, which to promote one alpha expression, which is in reach of those codons. If one alpha is a hypoxia inducible factor one alpha important transcription factor to promote cell proliferation and their hypoxia, often seen in cancer. This type of a codon specific translation has been implicated to promote specific protein oncogenes in different type of cancers. Another example I'd like to share is a trim tRNA, which instead of a single methylation on the ninth position in the D loop of tRNA. So, the patient that has a nonsense that mutation in trim tRNA to in the MbK, the patient have intellectual disability also presented on young onsite diabetic syndrome. If you look at the expression level of trim tRNA in reached in the brand added so measures really well with a disease phenotype. In 2020 this group I have identified there are more than 10 tRNA that modify by trim tRNA and missing trim tRNA the model and modification will lead to the confirmation stability alteration of tRNA initially medicine that could tune down translation. In the same year my group identified trim tRNA could interact with mRNA modifying enzymes to actually coordinate the methylation across our species to tune down translation. Because the patients show down regulation of trim tRNA so whether the disease phenotype is really because of the kind of modification here or could from other alternative mechanisms is still not entirely know. So I had to switch my gear to rebel zone are modifications so it's the modifications are mostly included during this complicated the concerted and the clinical important the biological event rebel zone biogenesis. In the description rebel zone Murray is under go a lot of a process things and always build it into the mature rebel zone for translation. Both the small subunit rebel zone are a ATS rebel zone Murray and the large subunit rebel zone Murray is all carry chemical modifications. Today, I like to draw your attention to the four enzymes that modify the ATS rebel zone Murray, because I like the enzymes are modified based in the large subunit. These four enzymes are absolutely essential. No cell can survive through knockout this four enzymes that modify the base of the small subunit rebel zone Murray. They are implicated in different types of disease. For instance, the MG one the method transfer is as the implicated in bow and cardinium syndrome, which is a rare genetic disorder that can lead to devastating consequences. Most of the patient will not survive through infancy. So the whole exome sequencing suggests the MG one carries as part of the 86 to glycine mutation in this conserved as part of the 86 site. This mutation will lead to a demonization decay of MG one. MG one has two roles. Cytolithically, it modified at the single methylation in this complex structure in 18 as a rebel zone Murray, a non-cytolithically, it participated in 18 as a processing and the small subunit assembly, because the patients show low level of MG one. So whether the disease is because of the methylation in rebel zone Murray or because of defects in small subunit assembly. The second modification on rebel zone, but this is what we know is more about the modification. So this complex modifies the two primal methyl in rebel zone Murray. The methylation that are required for the translation of internal rebel zone and inside the transcripts that important for hematopoiesis without the methylation this group of MR a the translation will be tuned down to inhabit a hematopoiesis that has been implicated in different types of leukemia. One last thing is about our MRA modification. I yesterday we talked about studio so today's I'm 6a. I'm 6a in MRA was first discovered in 1974 1975 for a while we think this is a static modification. Until early 2010, the discovery of the demethylase enzymes as shown. I'm 6a is under this reversible regulation that anybody was available 1987, but until 2012 two groups identified a transcript on why the location of I'm 6a using antibody assisted approaches. So how I'm 6a perform the biological function because of the reader pertains the reader pertains to have a higher affinity to the methylated than the emethylated transcripts to influence the splicing nuclear nucleus to set a plasma translocation enhance translation decay. Finally, this is an oversimplified model, but this is one of the earliest the generic models show the, how the reader pertains exactly to the biological function later on there are more reader persons discovered and the different types of mechanism proposed that this is a rapidly growing field with more things to be discovered. So, I mean to tell us three, the I'm 6a method transfer is having implicated in many types of malignancy, making it attractive drop targets. In that scientist went ahead and discovered, developed the inhibitors, they could inhibit I mean to us three method transfer activity with a pretty good at I see 50. They have tested the efficacy of the inhibitors in patient derived ML models and ML xenografts. Actually, this particular inhibitor already been FDA approved and enter trial one let's see how soon it could enter the market to start to help patients. I like to end my part by posting a few questions that I and my group of things about and for the committee to consider. For the TRA and the rebel zone are a modifications, because of the years of study we know better than I'm our modification about their occupancy in a given site and the precise location. However, when we talk about those modifying enzymes in disease. Are we really talking about the modifications or the enzyme by itself. So I think in the future studies, we could consider more of using the political in active variance to rescue the phenotype to see how much it's actually coming from the modifications. For the MRA side, at least the five six say we know it's more about the modification itself, what we need to improve is the single nuclear type of secret resolution mapping, and also the occupancy at the given site to quantitatively detect that. I'm really curious about how much the methylation in the given site will be tuned up and down to give a phenotype to give a pathological impact. Another thing I think it's a time to start to push this angle the coordination of modifications across our species. There's three major species support translation with the modification on them concurrently. So why is there the modifications in message are a rebel zone or a could the synergistically impact translation. When my group started to see some exciting evidence, we show them in terms of could have physically interact with each other to influence and coordinate the methylation in different species. In a more direct case, we observe the synergy between the TRI anti codon loop region and I'm six say to promote translation. Another thing is iron modification as a tool for disease diagnosis and and treatment is everyone still to use the only one, the best one for this purpose, or we could harness the power of a different iron modifications are, or even in a combinatorial way to treat the disease. Oh, yes. I like to send my like to send my group and your attention. Matthew that was very informative. We're going to move on to Dylan Simon and take questions at the end. So, Dylan Simon currently serves as the director of policy for the every life foundation for rare diseases. Since joining the every life foundation he has focused on newborn screening and diagnostic policy issues as well as annual appropriations efforts for the rare disease community. And Dylan. Perfect. Thank you all for having me. As you heard, I am policy. I'm not science person. So I'm here to talk about the impact on the rare disease community, speaking from a high level in terms of how this research can impact community as a whole. I want to talk a little bit about the organization just so a little more about me and where I come from. So the every life foundation is a DC based nonprofit nonprofit organization. Let's focus on empowering the rare disease patient community to advocate for impactful science and legislation. Really at the end of the day we're looking for how can we advance equity development of an access to life saving diagnosis treatments and cures. And I think it's really important to note. We're good. There we go. Okay. And then the last slide talked a little bit about how we empower the patients and so for us. We do a fair amount of advocating on our own, but at the core part of our foundation is how can we help speak a patient speak for themselves. How can we ensure the fact that they're out there. And having the ability to impact advocacy impact research because within rare disease. Such a key part of that journey is the fact that you are going from doctor doctor with people who don't necessarily know exactly what your symptoms mean or what condition you have and because all these things are rare as well as ultra rare. Many that they meet with have not even heard of the condition that they have. And so a lot of what we do is make sure that we can empower patients to. We can speak for themselves have to provide training so that they are aware of how to advocate around policy and issues that are important to them. A little bit background on rare diseases themselves for those who do not know. So they're approximately 10,000 rare diseases within the United States impacting approximately 30 million. You can see there there's a little conflict in the numbers, which is my apologies, but that is new. So previously have been 7000 we're in process of thing numbers to say 10,000. And one of the important aspects to highlight is that only 5% of rare diseases have an FDA approved treatment. And so that is really the key part of today's story. One of those treatments are either enzyme replacement therapies, you have a growing number of gene therapies, 70 to 80% of rare diseases are genetic origin. But the key part I want to highlight that is because rare disease patients are looking for treatments and cures wherever they can find them. They can't afford to be picky. They're quite often looking for how can we manage our symptoms. In the rare disease community is what is known as patient focused drug development. In sharing the fact that you're actually talking to the individuals with the rare diseases to understand what treatments and what treatments they want and what are the symptoms that they want alleviated. There's a great story I actually heard last week from a from an advocate who was discussing how when they were part of a clinical trial. A. They were. The trial was listing negative symptoms as a part of potentially into not approve a drug. However, they had not spoken to enough of members of that community realize that those negative symptoms. Which is a symptom of the rare disease. And so what they were assuming was a negative impact of the potential treatment was actually just the natural history of the rare disease itself. And so. We, there are family research national history studies to try to ensure that doesn't happen. But it really gets to the idea within the right diseases that. How can we ensure the fact that we're getting treatments to patients and treatments they need. And it's also from the economic side, quite important because as you can see here, when looking at just 379 of the more than 10,000 rate diseases. The economic burden was almost $1 trillion in 2019 alone. And so what you see here is that when you're looking even at a small number and you cannot generalize across all the rare diseases that this does have a high prevalence and high economic burden for all rates. For the full health care economy. And you can see here how it's broken down and you, the first would be direct medical costs. And so here's what we're talking in terms of. Inpatient visits drug pricing. What we think of in terms of medical costs and within the conversation we're talking about today. Inpatient visits is what I want to discuss most and so within that 418 billion they see there for medical costs, the highest percentage, the highest cost within that is in patient visits so it's. Those are the rare disease community having a onset of symptoms that they cannot care for at home so they have to go to the doctor they have to go to the ER they have to get taken care of. Obviously, the theory would be if you can see an increase in treatments that that costs would then come down. Second, when you're looking at productivity loss. And so these will be called indirect costs. So that is what happens if. The person with the rare disease can no longer work. In addition, what happens if it is a child, and the parent has to also leave the workforce and so that's where you're looking at productivity losses. And then the non medical and uncovered so what happens if you need to build a wheelchair ramp for the home aspects like that and so all of that we want to get to take the full picture because more often than not when you're talking about the costs. Of a rare disease or cost of a disease in general you're looking at the direct medical costs and they, as you can see are quite high. And, and do not mean belittle but you have to look at the full picture because it is not just is not just the cost of visiting doctors not just the cost of your drugs it is the cost of making sure that your home works for the rare disease patient is is the cost of can you still. Can you still your parents to go to work and so we want to look at the full cost and the reason I want to bring this up is because at the end of the day. Raiders these patients when talking about treatments just want new treatments they want access to these ideas they want access to new clinical trials is a big effort within the community to how can we improve access to clinical trials. Through telehealth and so it's how do we get this idea of decentralized clinical trials and so we're talking a lot obviously about our name applications today. And I think for me really the key point with the highlight is that more research is needed and within the rate of these community. That's all that anybody really cares about is how can we get this research done whether it is through natural history studies whether it is through gene therapies. Whether it is through any modifications. And so I want to highlight a couple stories at the end just to really bring this back to the patients. And so. The story you see here is from Lisa G who is a rare mom to Samuel who had who had a rare blood disorder. Samuel passed away at the age of two. So I'll make sure I get the story correct. So I'm going to reach my notes now. From rare blood disorder known as Langerhain cell. It's a senior. He was diagnosed at 10 months, which both these stories had early diagnoses, which is, which is great. However, in the ratings community. There's no something that was in diagnostic odyssey. It takes on average 6.3 years from onset of first symptom to official diagnosis. I know we're talking a little bit more about treatments today. But I think it's always worth bringing up the diagnostic odyssey because it plays such a large role within the rare disease community. And so Lisa and her and her family spent the next 16 months. They ended out of hospitals trying a myriad of treatments to try to find what would work for Samuel. Unfortunately, none of them worked. And he passed away at the age of 2. And unfortunately, in that 2 years who's missed out on a lot of the normal aspects of life because he was immunocompromised. And so when asked what issues were important to her. Lisa highlighted that more research is needed. It wasn't necessarily that. I need a specific treatment or I need, I needed more education or or I needed better doctors. It was, I need more research. And you see here that after Samuel passed away specific genus identified. That's provided opportunity to target the diseases and more effective life saving treatment. So while there's still no cause, she highlights the fact that the more research that occurs and the more funding that is going into this, the better these stories can come out. And so, and the reason I want to highlight the story is it wasn't. Lisa doesn't say we need more gene therapy research. Lisa didn't say we need more research into diagnosis. We just need more research. So within the radio disease community that is so important to talk about the need for rare disease research also looking at research that covers. Broad aspects of the rare disease community and cats. Those fantastic work looking at this as the idea of looking at multiple diseases as a whole. How can you create platform technology that will impact multiple rare diseases or rare disease families as opposed to looking individual diseases. And so, you can see Lisa's quote here on the slide talking about really luminous aspect of how we just need that more research. And so you see at the bottom of law doctors want to research the more common disease because of what can mean for the career. The rest of the community appears to need more researchers. I think that's such a great point is when doing the research, how can understanding how can it impact select rare diseases. And lastly, I want to talk about Nikki S. Nikki S was a rare mom whose son had TBC K syndrome, which is a rare neuro genetic disorder. And son was diagnosed at one month with again the TPC K. And met with many specialists when trying to determine what exactly the condition was. And she discussed when trying to access a therapy. The challenges that surround a therapist for TBC case syndrome was are more associated with the lack of information and research as TBC case syndrome is so new and research funding remains in early stages. There's only emerging research that supports these branching new assets as potential drug therapy to overcome some of the physical challenges patient face. So I want to highlight this for a few reasons. First is talking about how there is still lack of information the need for research such as what we're talking about today to really round out the story have a better understanding of what exactly is occurring within these rare diseases. But again, it highlights at the end. What I talked about the beginning in terms of physical challenges patients face as you can see here she's not talking about a cure she's talking about how can I just help to improve some of these challenges that my son is facing. And so really the goal for today and have to answer questions is really just again highlighting the fact that research like this can have a huge impact across the full rate of these community. They're they're just searching for new treatments and new cures to help alleviate symptoms potentially find some cures, but really just have that research have a better understanding of what their condition is, and has the best way to go about treating it. I think so much and happy answer questions. Thank you. Thank you Dylan we're going to take questions at the end of all three speakers so I'm going to be introducing next Philip yes key who's online just like I am. Are you there you are I see you. So I'm so so excited that you're here let me introduce you first. Philip yes key is the United mitochondrial disease Foundation science and Alliance officer that United mitochondrial disease Foundation is abbreviated UMDF. Dr. yes key has been active in the mitochondrial disease community since 2004 versus a parent of an affected child, then as a trustee of the UMDF Foundation from 2010 to 2012. In 2013 he accepted the position of UMDF science and Alliance officer responsible for leading the Foundation's research mission and managing all scientific and business development efforts related to improve diagnosis, development of treatments and cures and patient care. So, Dr. yes key go ahead. Thank you Susan thank you organizers for inviting me to be a part of this workshop I'm really thrilled to be here and you know I'm going to try to build off some of the comments that Dylan just provided you more general around rare disease therapeutic development and some of the case studies he did but I'm going to focus on mitochondrial diseases. And as you heard I have a very personal connection to this and my first daughter was diagnosed with the mitochondrial disease back in 2004. I'm a synthetic organic chemist PhD, when they said mitochondrial disease, I had no idea there were diseases of mitochondria and I suspect many of you may be in the same boat. So I'd like to give a little bit of background on mitochondrial diseases and then talk about how we can develop the patient voice of mitochondrial disease and the role that plays in effective therapeutic development. So, you know, as Dylan said, many of these rare diseases are really collections of ultra rare diseases so there's a kind of collective prevalence if you will for mitochondrial diseases of approximately one in 5000 that's the best epidemiology we have at this point, maybe even more common in that but it fits squarely in the rare disease space. Very high level terms, these are diseases of energy of course the primary role of mitochondria is to generate the energy that ourselves require around 90% of the energy is generated through mitochondria and the electron transport chain. And we'll talk a little bit more about the genetics of that here just in a second. Maybe an idea, these are diseases that can affect young can affect all adult onset is not uncommon at all. They may be mild and more chronic in nature or they can be severe and all the way towards a fail. So there's this broad spectrum of presentation. And just that encompasses mitochondrial diseases that really touches upon all the demographics inside of our society. I think importantly, no FDA approved therapies for mitochondrial disease and that really speaks to the large unmet medical systems that we're going to come back to as I continue. So from a phenotypic point of view, because all the major organ systems require energy to function. They really can impact a brain, a heart, the lungs, the skeletal system, muscle system. So you see sort of the range of symptoms that can present and have some kind of underlying mitochondrial pathology associated with it. So the genetics of course mitochondria organelles embedded inside the cell, the proteome of mitochondria is approximately 1100 unique proteins, only 13 of those proteins are actually encoded by mitochondrial DNA so the mitochondria have their own little genome that is kind of what I had to remind myself back in 2004. But of course the nuclear DNA does the bulk of the work so we have this symbiotic relationship between a mitochondria mitochondrial DNA and the nuclear DNA to ensure that the cell works well has all the energy that it requires. So the mitochondria the difference of course in the genome is we have many copies of the mitochondria inside of there so circular DNA and mitochondrial DNA. It can be hundreds of copies, thousands of copies in each cell just depending on the energetic needs of that tissue type. Whenever we have some kind of insult that takes place, whether it's in the nuclear DNA or the mitochondrial DNA that can cause aberrations and mutations inside of it. And for mitochondrial DNA because we have so many copies of it inside the cells. Now you have this concept of heteroplasmy we're only a portion of the DNA is mutated so the ratio of wild type to mutate is very important because there's typically some kind of pathogenic threshold that has to be crossed there. So this also affords an opportunity right to think about therapeutic challenges and how we get opportunities and approaches, how we can shift the heteroplasmy back to a sub pathogenic threshold. So, in summary, it's complicated when we talk about mitochondrial disease there's nothing simple about it, really both clinically and genetically, a complex set of diseases and this of course, you know, leads to a challenging diagnostic scenario and again Dylan spoke to diagnostic to see historically clinically patients are seen by doctors there's some assessment of the phenotype of the symptoms are showing that may lead to a suspicion clinically of mitochondrial disease, some blood tests can be done. Historically biopsies played a large role of just removing a piece of muscle and then looking at the bio energetics associated with that. And you can see some of the lab work that was done. But over time genetics has really become the primary way to diagnose these inherited disorders which totally makes sense. And over time, the genetics have advanced to the point now with genetic testing, where it moved from PCR through to very focused panels, but it was really this you know this diagnostic odyssey that begin Dylan spoke to is very much the case for mitochondrial disease patients. You know they're seeing a lot of different doctors and 70% are still receiving muscle biopsies this number is dropping rapidly though, and a major reason for that of course was the advent of whole exome sequencing and early 2010 kind of time frame, which now allows us to look much more broadly across all the proteins associated with mitochondrial disease. And I think what this little chart shows you is there's a lot of different pathways that are connected between nuclear DNA and the mitochondrial DNA including some transcription pathways that will bring RNA modification into question as well as a possible way to address mitochondrial disease. All of this is, you know, a huge more information rich approach way to get the diagnosis but it's genetically heterogeneous. Importantly though, by using whole exome sequencing. Now, our best estimate is roughly 60% of cases are solved using whole exome sequencing and another 30% of referred meeting. It's a mitochondrial disease but it was another genetic disorder, and that helps advance the patient's genetic, excuse me diagnostic odyssey as well. So I just briefly want to focus on neurologic symptoms and give you some example of how we try to develop the patient voice around this sub population of mitochondrial disease patients. And as you can see, and as I showed on the first slide, a broad range of symptoms that can be associated with neurological problems from stroke all the way through to ocular problems or hearing problems and cephalopathy and cephalopathy and cephalopathy and cephalopathy are also a really important component of mitochondrial disease as well. And when you take all of this in aggregate major unmet medical need, it will require a multitude of innovative approaches to solve these problems. A mitochondrial disease is not going to be the silver bullet disease that can be cured and taken away by a single approach. So it's really important that we're looking at this from a lot of different ways. We're going to go a little deeper on on something that Dylan mentioned around patient focused drug development and the FDA deserves a lot of credit for starting the patient focused drug development effort, again back in like a 2012 2013 time frame, where there's a recognition that in addition to the typical efficacy and safety data that the FDA may receive from any sponsor, it would be really important to get complimentary data from the patients to better understand what's most important to them, understand the burden of the disease, but also understand what do idealized treatments look like, as these NDAs come in so it's complimentary material to be used with that. And so successful, they couldn't fund it all themselves and they created the externally led patient focused drug development meeting, which is really the opportunity for organizations like UNDF to partner with the FDA put a meeting together, invite patients, caregivers clinicians and importantly FDA personnel, so that the agency personnel have the chance to meet and hear these patients. There's another perspective as they're reviewing these applications for for therapies. So UMDF co hosted this meeting, we called it energy in action back in 2019 we did it in Silver Springs right in the shadow of FDA. There were almost 50 agency personnel that came and participated which was really, really great across a lot of different review divisions and it really gave them insight into it. Of course we create a lot of data out of a meeting like this so it's recorded is transcribed. And we also wrote up a very important report called the voice of the patient report, and that summarizes that this is a lasting document then that is available publicly for industry scientists anyone to use to better understand the patient voice and for something as complex as mitochondrial disease. We really had to think about what can we cover in one day. We broke it down into two populations. We first we address the adults with myopathy so muscle based systems, a symptoms, but in the afternoon we focused on these pediatric patients with neurological symptoms we syndrome is the most common form of pediatric mitochondrial disease so I'd like to talk about that, that that neurological component just a little bit more to give you an idea of what we drew through discussion through testimony through panel discussion, you know, understanding the sort of range of symptoms that they're experiencing and you see a lot of them listed here, really gives you an idea that mitochondrial disease patients are not typically affected in one single way but rather in a in a range of ways the actual average has been published as 16 symptoms for mitochondrial disease patient. They have their ability to go about their life they're daily living. And so, loss and gross motor function and fine motor activities you know inhibits their ability to go to school go to go to work. These are all quality of life issues for my mitochondrial disease patients. There are also real social emotional and economic consequences again Dylan provided some background information. Our patient community is no different from any of the other rare disease communities in that it generates a significant amount of frustration and feelings of social isolation. There are important perspectives to capture and consider, as we're advancing therapeutics. Are we addressing the most important concerns. So, one of the first thing to do is to acknowledge the current state of affairs is not good enough, only 23% of patients through the use of vitamins and supplements as standard of care feel that these this type of lifestyle really affords them a good quality of life. So, again, a real need for improved therapeutics to improve the lives of these patients. What should these patients, excuse me these therapeutics look like well the most important symptoms to the patients we capture here and in this case, they do want to reduce fatigue and muscle weakness they want to gain function prolonging life obviously in a pediatric population, really important. When we get to the, you know what is it that would make a patient feel really excited about the developments taking place. Side effects are always occur, of course a concern they want to know how long go beyond the treatment, but that very last bullet point I think is an important one that was we teased out through this patient focused drug development meeting. These patients are willing to accept a certain amount of risk for a specified benefit. And that to me is a clear message to the FDA that we need increased flexibility within the regulatory framework. I'm not to say that anything unsafe should be prescribed to the patients approved and able to be prescribed, but these patients are willing to accept risk in order to get a benefit that aligns with their most highest prioritize symptoms. I'll wrap up my comments here quickly. Of course there's a large unmet medical need with no proof therapies, the diagnostic journey, similar to what Dylan said is can be multi year long complex, they're seeing a lot of doctors they're they're experiencing a lot of symptoms. I think, importantly, I hope my comments have reinforced for you the important of getting the patient's perspective inside of therapeutic development. And there are many ways to do it the patient focused on development meeting and the patient voice of patient report is only one example. We also run patient populated registries which allow us to collect data and and build this perspective from the patient perspective. There's also patient led research we're part of consortia where the patients try to identify what's the most important what are the biggest gaps that will address our largest concerns and this has to be done in close collaboration with the scientific community, but it just creates a different perspective in terms of what direction we're going to go based on what the patients have identified as top priorities. I cannot emphasize enough that in general across rare disease, we will not achieve success in developing therapeutics without increased regulatory flexibility I know this is very important that every life foundation, certainly important to the United Country of Disease Foundation. We strongly encourage our friends at the FDA to use accelerated approvals and bring these drugs to market and understand that patients are willing to accept a certain amount of risk for the potential of a benefit and improving their lives. And then lastly, I just want to mention that for, I think many of the scientists in this room. Nothing replaces the opportunity to go to a scientific conference and have patients families there with you co attending UMDF has done this for almost 25 years and I've put a link to our our conference which has a very robust scientific program, but also a patient track where we always look for opportunities to intermingle all those stakeholders and consistently we always hear from the scientific community. I left with a much better understanding of the patient perspective, and I think that's our goal and all of this work. So again, thank you and look forward to your questions. Thank you, Phillip. Alrighty, so thank you to our three speakers and now we'll move on to questions from the committee and from the audience online reminder that you can put your questions in the, in a link that you can find in the chat. I'd like to start with the very first question for all three of you and anyone can can take it is that all of you have advocated for increased research on rare diseases, some of which can be attributed to problems in our name modification. Who do you, and as someone who does work on rare diseases. I know that my research doesn't go forward without funding. So what do you think should be funding research on rare diseases, and how do you advocate that they should be doing this. So I don't know who would like to take the first question Philip, I can see Phillip on the screen so I'm going to. Yeah, no I'm happy to share a couple of thoughts on that so you know UMDF is a, you know, sort of a soup in that patient advocacy group right where we we do span from research to education to support to awareness. And so patient advocacy groups can play a really important role in seeding the research necessary to develop these therapeutics and achieve our overall mission goals and so you know UMDF has made over $15 million worth of research grants available to the community. But those are always meant to be seed grants, generate sufficient preliminary data that hopefully allows the researchers the investigators to go and leverage that into a much larger R01 from someone like the NIH or the Department of Defense NSF there are you know a variety of larger pools of funds so you know I think it starts with patient advocacy and that you they can see the market but it really you really depend on the large funding agencies to drive it. And Dylan do you have some thoughts on that. Yeah, I can hop in. So the main place is going to be at the NIH, which I don't think surprising to really anybody. This is really the home of rare disease research at the NIH, all institutes conduct their own, but you might get more to see specific. Once you get into individual institutes, whereas NCATS houses their own division of actually essentially gave it a promotion. A couple of about a year and a half ago from an office to a division of rare diseases research innovation. And the reason I want to highlight that is because that's the NCATS is where you're seeing cross platform research and so one one example is what I call it what is they call pay GT which is platform vector gene therapy. So looking at how can that technology be utilized to treat multiple rare diseases, instead of looking at individuals and so I'd say NCATS is the best place to advocate for also like to note we're in budget season so I have to say this as a policy person. NIH receives about a 1.5% bump from the present budget which is typically lower than what we see but also present budget tends to be lower than what Congress eventually does do an increase for NIH. But the reason I want to flag it is because NCATS receive flat funding so all the full NIH got 1.5% NCATS got flat funding. And so, if you are advocates and do speak to your senators or congressmen please let them know to support NCATS research. Outside of that I did want to flag that FDA also does some research here this is more on a little more on the clinical trials side, but I think it's important to note they have what's known as the orphan products grants program. And they have 2 funding streams within that 1 would be natural history studies, which I talked a little bit about as well as clinical trials. And they also have a new grant program out that kind of runs tangentially to that which will look at specifically neurodegenerative disorders. Oh, sorry, go ahead. Kathy, any thoughts. Okay, so I probably can speak from a researcher's perspective. I think the bigger contribution part could come from age because it's more longer sustainable funding resource for a researcher, but we do really appreciate the money that coming from foundation because we know it's actually donated by patients at least for myself. Whenever I spend in the foundation dollars I feel I'm responsible for a cure. Yeah, no I think I think that's really important for those of us for those of us who do that that that perspective of researchers very important. So just to follow up on that, now that we have a kind of your ideas on who should be funding it. How do you, how do you advocate for something that's rare. So how do you make that argument, you guys must be up against this all the time help us. I can start again. If you like, you know so. Yes, this is always a challenge. You know, again, I spoke to the fact that you know for for many this word mitochondria is still be a very strange to them not so much in the scientific community. There have been tremendous advances right and understanding mitochondria biology to the point now where we know these rare inherited disorders are a perfect starting point for how to improve mitochondrial function. It helps these rare patients, but because mitochondrial health is central to the human health condition it really has the chance to impact literally every person on the face of the earth right with healthy agent. So I think that mitochondria eat well exercise you know no big surprises, but it really makes a huge difference and it's been demonstrated to be an important part of it. So I think we're fortunate in that we do have a connection that that extends out, but for many rare diseases, it doesn't right it's just around that that very small patient community it doesn't make it any less important. It's just a different situation. Anything to add. So coming from the broad rare disease community. We always say rare is not that rare, because you have one in 10 are impacted by rare disease. Do speaking with a collective voice telling your story is always a key part for us. We just had our fly in week. Two weeks ago at this point in which advocates are meeting with representatives to tell their story. And it's in those meetings that can be so impactful so having the numbers is always great. There's a reason that we did that economic rare rare economic burn of rare disease study, having that number is really helpful. We're also looking at the cost of the diagnostic odyssey. And so having those numbers, I don't want to belittle that having those is can be extremely important, but pairing that with a story is what is needed because numbers will get forgotten, but a story will not. And so it's, it's personalized in the story a little bit can can be so important specifically within rare diseases because is thought to be so uncommon even though it is not. Anything to add Kathy or should we move on. We can move on. Any any questions from the audience. We have one question in the room and I think we need to move on, because we're a little behind schedule but one good. Yes, so my question is for Philip, and I'm clearly part of this enterprise of sequencing and mapping mapping modifications. It is a reality that we need to talk about scale, and the scale means different issues different organisms. So from the point of view RNA types in particularly mitochondrial diseases. In your view, would you comment on which are the ones that are most affected by modifications. Well, yeah, you know, hard to give a short answer to a question like that. You know, I focused on the neurological symptoms and you know and brain tissue but it also presents the largest challenge right and being able to cross the blood brain barrier and affect change inside neuronal tissue. You know, the greatest amount of clinical activity in the space is in the muscular system. And so, you know, if you look at the symptoms that patients report many of them are associated with muscle fatigue and weakness, and this impacts their quality of life in a myriad of ways. So, I think the ability to, you know, make modifications that are meaningful inside muscle tissue, maybe the entry point right for coming into the mitochondrial disease space and hopefully then your heart tissue and brain tissue. You will also be a part of the future developments. Great. So Steven you're taking it's all it's to you now. Great thank you so much Susan and thanks to all our panelists are going to move forward now, because we have a little behind schedule but we want to keep moving and we have a little bit of time. We have a little bit of time left for the end to play with so Kate, I'm going to bring up sorry I'm going to bring up one of our committee members Kate Meyer, who's going to walk us through what to do for breakout groups. Great, thanks. So, right. Next we're going to branch out into some breakout groups and discuss some important questions. So for those of you who are online you should be put into a breakout room room automatically and for those of you who are here in person. If you didn't grab one of these sheets at the table out there. Please do so after I'm done talking here in a few minutes because it has everyone's breakout group assignments. So we're going to talk very long I'll just state the six main questions that we're going to tackle as a group so the first is establishing the near and long term goals for mapping and sequencing RNA modifications. The second is understanding educational and workforce needs. The third is understanding the impact of epitranscriptomics on health and medicine. The fourth big question is establishing ideas metrics and standards for database and infrastructure needs. The fifth topic will be establishing metrics of success in the field of mapping and sequencing or any modifications. And then lastly determining the importance of the field within the larger landscape of the life and medical sciences. So for those of you who are going to who are in person I'll just make a quick note that if you're in groups one, two or three, you're going to be here in this room discussing. In groups four or five you'll be right next door in 118. And those of you in group six are going to be just across the hall in 114 and there will be someone outside to direct you. So with that I'll let everyone kind of go into their breakout groups I will say, please designate one person who can be your representative because we're going to reconvene at 1045. So if you have the representative from each group, tell us about how they address the main questions under their, under their primary topic. So I'm good with the, the in person breakout group for the first topic of establishing the near and long term goals for mapping and sequencing RNA chemical modifications. So that person can just come up to the mic and you know I would say just two minutes or less just summarize what you guys talked about if you want to address each of the specific questions that's great if you want to just more generally describe what you guys discussed that works to Thanks Kate. So the first group we had a lot of great discussion I think from the short term perspective one of the key outcomes, if we're going to be able to do mapping and sequencing is the technology needs to be improved. There's no doubt the technology is not there to do this at the scale that folks want to do it with the accuracy and with the end goal of having useful Some other ideas that were brought up in terms of short term focus is how to bring new users into this area. What can we do to incentivize additional research and researchers who are interested in this. One of the ideas that were brought up by the group or do you think about starting to do crowdsourcing do you do a little game theory so you make it more competition based different ways a user curated or a user involved database to do this. I think the long term goal is how do you have a source of truth that is out there that everybody can go to look at and say regardless of whatever transcript I'm looking at. Here's where the modifications are for this particular system whether it's this cell under these environmental conditions what it might be that would be obviously the holy grail or unicorn, maybe as we think about it. Let me check my notes. I think some of the barriers that we identified it are access to standards. So what are those standards barriers also involve, you know, what are all the right approaches, including the reagents, antibodies, the purification, the samples that everyone would have. So barriers on the data side. So how do we access all this data who stores it who's going to be responsible for quality control came up more than once in the conversation to understand how this is and the need for some SOPs so that everybody is understanding how things are being done. Some of the, the thoughts that were shared is, you know, one of the ideas and how do you start organizing to move towards these goals and get this is do you think of consortia that are based on that are charged with specific modifications and specific types of RNA, and they sort of help define and set what their goal should be within those groups, or do you do it a little bit broader and you let the whole community be involved. I know my group's all over is there anything else we miss. I miss. That's great. Thank you. I think, in the interest of time I won't add further comment but I think that's fantastic touched on a lot of things we heard about yesterday but also some new ideas for how to address some of these hurdles. So let's go now to online group one addressing this question. Do we have a representative from the breakout group one online. So we don't know what numbers we are so you should. Josh Burtick is. Oh, yes. John, why don't you just give it a go. Sorry, I lost track of the number there. Yeah, we heard from people who worked in mitochondria, someone working in mitochondrial disease. And in that case, it seemed that mass spec was the most useful way to find modifications in tRNA, and also heard someone working in ribosomal. RNA finding that nanopore and other mass spec was not as useful. So it seems like there's complimentary structures, strengths for different technologies, and also mentioned was that in embryogenesis and many other things there's cell type specific things which so addressing that would be important. And also the need for standards is clearly there. And I think that's the biggest things that I heard fantastic. Yeah. Thank you very much I think this is great. Okay, now we can move on to to group two. In person group to understand an educational and workforce needs. Hi. Yeah, so I was in group two, and we talked primarily, my name is David, and we talked primarily about putting emphasis on undergraduate training to meet this critical need to increase the number of researchers sort of working in this field. And we talked about sort of jumping off of Wendy Gilbert's point from yesterday about the need for more computational training for students more focus in the biological sciences at the undergraduate level. So we sort of rift on that for a little bit and sort of came to one possible proposal, which is, or if there was some pot of money to devoted to this, this need, we could sort of have institutions apply to get funding to host, you know, sort of one or sort of get students sort of off, you know, off to some sort of start to learn how to do computational biology, and the context of maybe analysis of, you know, results involving studies of our modifications. And this could be built. Each institution doesn't have to sort of build this workshop from scratch they could use existing workshops like at Cold Spring Harbor laboratories and other places sort of as templates to design their own. That was one idea and that we this would be executed in a way that would be mindful and sort of inclusive of sort of many sorts of students including those that historically have not been in this field or you know in the biological or computational sciences and so, you know, for example, if a student had to sort of leave their, you know summer internship for a couple weeks, or couldn't work for a couple weeks that they would actually be paid to do one of these workshops and this would be sort of an investment in the future of the field to having more diverse groups. The other thing we talked about was, you know, to understand a lot of these talks that we've heard, you have to have some understanding of chemistry it's RNA chemical modification so the need to promote this research and sort of undergraduate curricula was discussed as well and sort of how do we sort of get chemists excited about even those that are not in our field excited about teaching their students sort of at the undergraduate level about the sort of questions and that you know you can that we're excited about but they have no idea about and may never know until it's too late so we didn't come to any specific recommendations I think for that and we discussed the challenges of changing curricula, for example, you know, adding biology, you know, telling biology curriculum committees to replace a year of something else with a year of computational training, but for chemistry I think even just sort of introducing the idea of RNA chemical modifications in coursework would even be pretty helpful. And I think that was most of what we talked about. That's great thank you yeah I love the idea of having workshops and thinking about how we can introduce this topic as early as possible to our future workforce I think that's really great. Okay, so we'll move on to topic three understanding the impact of epitranscriptomics on health and disease so let's have our in person group three go first and then we'll do our online group three. Oh, I thought there was no online group to. Okay, well. All right, well, great idea. We'll do that we'll do all the in person and all the online. Okay, I'm tall pan. So our group three is in charge for understanding the impact of the transcriptomes and health and medicine. There were three questions. The first one is what is already known about how the epitranscriptome impacts different health conditions. So, so that I think we discussed, for instance, some modification enzymes they are over expressing cancers and they basically self become addicted to them. Why, because these enzymes and modifies for instance certain tears and these tears are needed to do translation regulation of certain uncle genes, such as a Mac and so on. And we also, okay, so then the then the question actually comes about is how do how about the actual modifications. How would these directly impact health and disease. And so, so those are of course, the best known for something like a mitochondria disease, for instance, the where you have mutations and mitochondria is actually leading to a missing modifications that actually real cause of dysfunction unless counter translation. And then the second is how has knowledge of the epitranscriptome be utilized treatment and diagnosis. So the treatment is, is again, I think the known drugs as far as I know that they target writer proteins for, for instance, may not ask for the MSA or writer proteins for the time SOG modifications but only targeting a specific cancer type. So that because these cancer types are specifically overexpressed them or doing something differently about about them, again, leasing to the misregulation of, say, oncogenic proteins and so on. And the one thing we discussed is actually, there is a, the billion dollar industry out there of small, small molecule drugs targeting the structure, missing our stability missing our localization. And for that, at least in a conference I went to last October. People talk about all these great things and yet there was no really consideration at least no data on what happened when the any of these areas are actually modified in vivo in cells and then so that is a consideration that perhaps mapping of the epitranscriptome at least in the RNAs were going to help out a great deal. And another thing is known for the actual modifications this are some classic examples will be snow RNA that actually impacting ribosome in our modifications that also turned out to matter of a lot in in cancer space. And then, finally, the other question what the information is missing, being able to fully understand how the epitranscriptome impacts health and medicine. So the first thing we talked about are the coordination of the three family of race, there's a ribosome and RIT and my son are modifications, and, and, and clearly, all three of them play a role in really making the proteins that going to matter for health and disease. And so, so, so, so yeah, so I think that's kind of a card. So I kind of information is currently. I mean, we know some specific cases but this, I don't make levels differently not there yet and, and it will be nice to to know these things. Fantastic yeah and I think you're touching on some really important points that haven't been talked about extensively yet in the meeting including, you know, small molecule targeting of RNAs and the impact of modifications and the things like, you know, targeted pseudo urulation and other things that can potentially be important therapeutically. Right. Right. Great. Julie selects we were grouped for on establishing standards and infrastructure so I've got four takeaway points. One is the universal need for standards I think we all recognize that both in terms of biological standards and the data standards. The second one is the need to consider the entire pipeline when thinking about these and establishing them from the synthetic oligos that we've talked about a lot at the meeting but also to the experimental protocols and a lot more discussions needed on data storage dissemination and even visualization. The third major point is the variety of challenges that this particular project will bring to light things about data format data storage and I think very importantly compute community buy in and compliance with whatever is developed. There's a lot of really interesting differences between DNA sequences and the RNA modifications that we want to map and sequence, both in terms of the complexity of data in the RNA realm. The accuracy of some of the techniques that are being developed to deliver us this data. These things are driving a value in actually storing and learning how to deal with the raw data so that if algorithms improve we can go back and still utilize this resource. There's also a really fascinating conceptual challenge with with the dynamic nature of these RNA modifications that really affects how much data is collected and stored it affects the metadata we need to have to describe this. The visualization of this data and the entire infrastructure for all of that. The fourth point is who is in charge of establishing all this. So we've got, and this is just a bunch of questions. So for development, you know we heard from NIST maybe they would play a role. There's also many roles for the community, both in terms of researchers and industry. Who's in charge of setting these standards, maybe NIST does some maybe IU pack even set some for example determine letter single letters for all of these bases. Who's in charge of maintaining this infrastructure and CBI came up as a possibility and who's in charge of policies to enforce this is this is an NIH thing with new data policies that just came online. Is this a journals thing, for example, new clay cast its research demanding deposition of the data. So, I'll leave it at that. Fantastic yeah these are all really important questions and I think really good to be thinking about them. Thanks. Okay, we'll do in person group five next group five yeah establishing metrics of success in the field of mapping and sequencing modifications. We're group five and we had two questions on establishing the metrics of success in the field of mapping and sequencing of RNA chemical modifications. The first question was, what are some of the advancement metrics for quantifying the progress in epitranscriptomics and so we had six kind of bullet points, the first one had kind of two layers to it. The first layer is standard workflow for sample prep processing for RNA modification essay that is adopted by an industry partner, we think that's really important to have something that's not just using in a lab. It's something that is actually, you know, a true sign of adoption is, or, you know, a true metric is advancement to the point where there is industrial, you know, one or two, preferably to and actually that's the second layer of that is to have competition within So for example, by, by self fight workflow for RNA seek or the Oxford platform for RNA seek libraries that that's a sign where things are now adoptable and usable by others. The second is a low cost technology for sequencing and mapping a particular modification routinely with established accuracy and bias so it's important that the accuracy and bias are fully established. The third is a good metric of success for Sorry, a good metric for success for one modification should be the drive for adoption for another modification so I think, you know, I guess, we kind of can repeat the same metrics for other modifications so it's kind of until we get through all And then the elevation of status of a particular modification to a clinically relevant marker. So if there's a particular either sequence that hosts a particular modification that is clinically relevant. I think that's a very strong metric of success for that Demonstration of a framework for integrating technologies for combinatorial modification detection. So this is more of a question of how do we integrate knowledge about, let's say we have assets for different modifications, how do we Integrate those into a unified Annotation And then Finally, either the organic or top down development of a framework for sharing data and ensuring reproducibility across users. And then the second question was how often should we be measuring success and who's responsible for measuring success. And for that we had We We thought that a standard type in age grant has the right mechanisms for for that. For example, an external review board and annual reviews by Program managers by you know internal reviews by by the participants in the in the consortium And in terms of the actual data and you know measuring the success in terms of the way the data is handled and managed. We brought up maybe maybe NIST for for managing some of the data and kind of curation of the standard data and you know maybe standard samples to And that's about all we had. Thank you. Fantastic. Thank you. Okay. Let's move along to the final in person group. Hi, Stacy Horner. So, oh, sorry, I can't read this. So our group was tasked with determining the importance of the field within the larger landscape and life and medical sciences. I thought many of these ideas were already covered by group three well so talk of a great example of our name modifications being important for disease. But I'm just going to summarize kind of our overall discussion, which is we don't know what we don't know. And so sequencing everything while we don't know what's going to be important it's really hard to know if you don't know what's there. So there's a clear need for that related to other groups we think there need to be clear standards and metrics and education so that everybody's doing the same thing. We have that map, then you can go further. So things that we know that our name modifications are important for regulating gene expression, you know that that includes translation. But the one group thing that our group also mentioned was that it could be important for drug design. If you think about what modern is doing some of the modifications we may not know how they're going to impact disease but they could impact function in a drug sequence for that reason they would be important to consider thinking about. And then the other case that we talked about is you know which RNA do you sequence. And it's the same thing we don't know what we don't know and so sequencing every RNA at some level is probably worthwhile, including RNAs from viruses bacteria and if you think you know our name modifications have clear roles in regulating different aspects of viral infection, antibiotics are novel antibiotics that could treat our name modifications bacteria could also be really important to consider so we kind of think, thinking across a broader tree of life would be important. Great thanks all fantastic points. Okay let's move on to our final two online groups and maybe in the interest of time. If you can let us know any of the new things that haven't been discussed already that you talked about. I guess we heard from breakout group two already so maybe let's start with breakout group one. Hi, I think I got posthumously elected to do this. Many of the things that we discussed on near term and long term goals have already been covered by other groups so try to make this short. We talked a lot about in the near term, needing a lot of new tools, specifically some examples were better nucleases so that RNA could be segmented into smaller oligos for mass spec analysis. And this would be nice to have a whole suite of these nucleus enzymes that would operate and very known sequences or motifs. And secondly, sort of plug and play ways to synthesize standards so that we could put any modification anywhere in a way that it can become a standard for for testing. For example, being able to develop methylation enzymes that could place methyl groups on either two prime hydroxyl ribose or on any base where you desire would be fantastic. Or other other ways to incorporate modifications so that doesn't have to be 100% incorporation and every 100% replacement as is done for vaccines. The third thing we talked about was basically facilities, could, could there be a movement toward having core facilities that could do RNA mapping so that not every lab has to be an expert in the techniques. We also talked about national facilities but I think local core facilities might make more sense. In terms of talking about long term goals, we agreed with other speakers at the workshop that the long term goals will emerge as we are able to study all the modifications and figure out what's really important. It's very hard to say that a priori which modifications will be the most important because it could be in interaction, or the confluence of two or more modifications that are actually what's particularly disease relevant. So, so we weren't really able to define long term goals yet. That's great. Perfect. Okay, we'll hear from our last online group. Hi, I'll report out for our group. I'm sorry I was in a group that had Todd, Shweta and Lee Roy so it was a really interesting group we had a few chemists a few biologists and also an SRO from NIH. So it was a really interesting conversation that we had one of the first major points that we sort of came across was that, you know, biologists and chemists need to need to talk more, and especially, you know, talking to our synthetic chemists about what our biology needs are and that still seems to be even though we talk about it a lot still seems to be a disconnect. And so that was one major point that we came across and I'll mention we kind of went off script a little bit and didn't talk about anything specifically related to what we were supposed to talk about but a lot of good stuff came out. Another thing that came up was that we can't really understand the big picture if we're not thinking about multiple modifications that could be on any given RNA molecule, how these work combinatorially and crosstalk within the same molecule and then also with other molecules. So that was an important thing that came up that I don't know if we necessarily heard but Cynthia kind of mentioned this in her, her report out as well kind of thinking about maybe multiple modifications that could be, you know, at play in any given situation. Another thing is, of course, everyone talks about standards but somebody brought up that maybe, you know, in her day to day work as a biologist working with mass spec, maybe the standard doesn't necessarily need to be at its most perfect stage like our NIST level standards, maybe we can all work together to work with what we have and kind of go from there. And then our very final discussion was sort of around funding and how we actually make sure that we're funding multiple efforts and not just, you know, one big effort and then kind of realizing later on will wait, we missed an opportunity so sort of talked about not specific ways but that there's a need to get funding out quickly, prioritize quickly and then kind of have regular check ins with the community and be able to discuss challenges and pivot very quickly in a different direction without losing the interest of the broader community and funding entities and so on. Yep, those are all great points thanks very much and I'm told that we did have one more online group as well so now we'll hear from our last online group. Hello. Thank you very quickly. Our group, I'm Jessica Seba Fisher and our group was tasked with talking about the standards and push register and database. I think one of the overall themes that we were talking about was that it is really too early to understand the complexity of potentially what a database can be and we really need a lot of more small projects to occur so we can identify the breadth of what we actually need. Some of the thoughts that we did come across and discuss were potentially that database definitely needs to be cross reference integrated. We're not only talking about, for instance, one field right we're talking about transcript domains are talking about mass spectrometry time up potentially structure. So you need to make sure that the database actually incorporates all of this right so it definitely needs to be highly complex. We also talked about the accessibility so need to make sure that it is accessible to all different types of biologists all different fields. So not only those who know how to do genomics or transcriptomics right bioinformatics also those who are basic biologists, including those at stages of all levels right we're talking about the younger generation who are highly computational already, but you know what about the other fields that may not be so computational. So just basically to make sure that everyone can access the database. Additionally, we also think we need to kind of step back. There are so many databases currently out there already that are extremely useful right in every single type of field. So how about just stopping and thinking about what we like from those databases and how we can incorporate what they are doing to utilize for a database and infrastructure that we have. Some of the other topics that we said were discussed there about storage issues taking the raw data how to compress it. Who's going to maintain it that's going to be extremely important is going to be a lot of information from a lot of different areas. So who exactly is going to be maintaining this and making sure that it is highly organized. Another thing we talked about, which was very important I think was also about the mass spectrometry right this is its own important really big field that's going to be in our any modification so not everyone can have a mass spectrometry in their lab. Right. So one thing we need to think about it's expensive is how can we actually outsource them maybe some type of incentives right utilize the current really great structures of the core facilities and things that are at all the different facilities, right and utilize those structures in order for us to get that type of information, and see what they have. When we did talk about the standards and metrics, you know, again, I, we all probably agree that is really too early to understand this may be disease specific and maybe the question you're asking that, you know, you may need to look at your standards. And we are really early so even simple things like antibodies are still being optimized of what we need to use with all the different pull down essays right. Do we need to look at stoichiometry doing look at the different modifications we don't even know what the percentages are of the modifications, you know, dependent on the disease type the state type you know even cell type. So of course, you know, do one of these standards actually need to just be biological outcome. Right. If the modification is important and it's causing some type of function, or you know some type of phenotype you know maybe that needs to be incorporated as well and that is going to actually be dependent on the question that you're asking. And lastly, I guess like I said the overall thing was that you know we are really early in understanding, you know what these modifications are doing. And so you know we really think we would benefit from having some small science being done and you know, learning from what you know these small groups are actually doing and learning from that to actually create this type of database infrastructure and standards as well. So great thanks very much, and thank you to all the breakout groups these are some fantastic ideas and lots of important things to think about. I'll turn over to Steven who can tell us about our break. Great, so we are quite behind. So I'm going to allow for a five minute break so, but please try to be back here 1125 sharp, so that we can make sure to get through all of the next session since there are a lot of folks who we still have to hear from, but thank you. Sarah. Great. Well, thanks everybody. My name is Julie sucks from Northwestern University and I'm very excited to be moderating this last session of the public portion of our workshop. We're going to be talking about emerging tools technologies methodologies and information that are not currently applied to the process of mapping and sequencing the every transcriptome at least not at the scale that we have been talking about, but there's a lot of potential opportunities to utilize these tools and ideas to to advance the field so we're going to be hearing about three types of tools and concepts. One is about the mechanisms of readers and writers that we've heard a little bit about. They might play into mapping these modifications. We're then going to hear about artificial intelligence techniques for sequence and RNA structure recognition and prediction. And then we're going to wrap up with some updates on cryo em for structural determination in this approach. So we've got three blocks of speakers, each speaker is going to speak for 10 minutes and at once the speakers are done in that section. We'll then open it up for questions as we've been doing. So, I'll kick off the first little sub section on mechanisms of RNA readers and writers by introducing down it's a Fujimori who's a professor of cellular and molecular pharmacology and pharmaceutical chemistry and associate director of the Quantitative Biosciences Institute at UCSF. Her lab combines organic chemistry and biochemical reconstitution to investigate enzymatic mechanisms and regulatory roles of post translational and post transcriptional modifications. Down it's a takeaway. Thank you so much for the introduction. Can you hear me okay and see my slides. Yes. Great. Thank you. Thank you for the opportunity to present in this workshop I enjoyed yesterday's sessions and I've learned a lot and look forward to today's program remain there of the today's program as well. What I'd like to tell you about is our applications of mechanism based strategies as well as structural methods to map our name modifications and at the center of this will be two enzymes that modify the bacterial ribosome. Many of you are aware bacterial ribosome is one of the major antibiotic targets targeted by 40% of clinically used antibiotics and functionally relevant sites for antibiotic bindings are those that are functionally relevant to the function of the ribosome as well in peptide bone formation. The majority of antibiotics that target ribosome bind in particular transfer a center region of the ribosome and the adjacent nascent peptide exit tunnel. The pathogens have evolved a number of mechanisms to inactivate antibiotics by modifying their binding site within the ribosome. Among these methylation predominates as a small modification that is sufficiently large to include antibiotic binding site without disrupting the function of the ribosome. Over the years, many metallating enzymes have been identified that decorate ribosomes through modifications in the peptidyl transfer a center and the nascent peptide tunnel as well as at the interface of the subunits. The enzyme that our lab has been investigating is chloramphenyl, chloramphenicul, fluorophenicul resistance enzyme or CFR, which confers resistance to eight classes of antibiotics including these five that are used clinically or in veterinary medicine. CFR as we abbreviate it modifies a conserved nucleotide, the A2503 and 23S ribosomal RNA, to introduce a modulation to C8 metal group of adenosine. Basically this nucleotide is pre-metallated by a housekeeping enzyme called RLMN, which introduces this conserved C2, sorry pointer, that introduces a conserved C2 modification that is not associated with resistance. Once hyper-metallated, this resulting ribosome is resistant to antibiotics. Here and RLMN are mechanistically related and our lab and others has investigated mechanisms by which methylation by these enzymes occurs. What is unique to this class of the enzyme is formation of a covalent intermediate in between the enzyme and RNA substrate. So let's just walk you quickly through the catalytic cycle. In the resting state, there are two conserved systems of which one gets pre-metallated and then activated by in a radical stem fashion to form this reactive thio-metalline intermediate. This thio-metalline forms the covalent intermediate in between the enzyme and the substrate, which is then resolved through the activity of the second conserved system. But our lab has demonstrated is that if we mutagenize the second resolved system into an alanine, we can stabilize this intermediate as shown here on a gel, we can enrich for it, we can isolate it by using a appropriate affinity tag on the enzyme and then characterize the nature of the bonding between enzyme and the substrate, which we use for mechanistic studies. But we have also expanded this to identification of substrates and sites of methylation by this class of enzymes. So briefly, the approach relies on incorporation of the alanine mutant, alanine mutant that cannot affect the release of the methylated RNA product on a flag-tagged enzyme in the background of E. coli lacking the enzyme. We use that flag tag then to enrich the RNA substrate. Protein is key to digest the enzyme, leaving a small polypeptide scar and then we obtain the enriched RNA substrates which are sequenced and identified through next generation sequencing. We've had tested several reverse transcriptases and for the purposes of this workshop I think it's important to highlight the need for reverse transcriptases with varying abilities to either co-stops or incorporate mismatches. We have used Tigard as a reverse transcriptase that reliably incorporates mismatches at the site of the polypeptide scar as a way to identify, to validate the substrates, but also identify sites of modification with nucleotide precision. So using this strategy we've been able to validate all of the non-substrates of E. coli aralamin. It's a rather promiscuous methylation enzyme in addition to 23S ribosomal RNA, it modifies several of the T RNAs. What we see is that these T RNA substrates are largely enriched and in the cases when they are not enriched such as this glutamine UG we were able to detect the substrates through mismatch incorporation. So another valuable use of mismatch incorporation in validation of RNA substrates. And with this I'd like to switch gear and tell you about structural methods that we applied for identification of modifications. In the past decade or more cryoEM has really revolutionized the field of structural biology including that of the ribosome. And in 2020 we've been able to obtain the cryoEM structure of the E. coli ribosome at a resolution of 2.15 angstroms. Now resolution of the structure is better in the core part of the ribosome in the peptidol transferase center and the resolution there reaches sub 2 angstroms which allowed us to map modifications. So the way this looks in practice is highlighted here. We've been able to use cryoEM density maps to model RNA modifications such as this methylation of U, C2 methylation of A2, 503, methylation on the ribosome of another nucleotide. And also to determine reliably positions of waters and metal ions based on their density and coordination spheres. In the process of this work, Iris Young, talented computational scientist in the Frazier lab has developed a tool that I think community will find quite useful that we call QPTXM or quantifying post transcriptional modifications tool. What this tool does is interrogates regions of the cryoEM map for automated detection of RNA modification. And what goes as an input into the tool is geometry of the modification. And what is at the appropriate angle based on the known hybridization of the atoms to which it is attached. Distance. So what is the distance of new density relative to the nearest heteroatom and does it decay as one goes away from the density. And of course the quality of the density, which restricts this tool to well resolved regions of the map. Iris has developed a QT plugin for interactive viewing of modifications that the tool calls for us, which is really, really critical given, like in many other methods, high false positive rate. Despite the high false positive rate, what we find the tool very useful for is calling modifications in an automated way that we can then manually inspect the sites predicted to be modifications to truly assess if modifications are there. And using that we've been able to, to convincingly annotate some of the modifications that are in law in abundance to geometries in well resolved regions of the map. So in just the last last minute. I'd like to highlight also our work on obtaining structure of hypermetallated ribosome, but the key problem was that the CFR is isolated from stapharius it's the enzyme that we and others do biochemistry it has really poor, relatively poor ability to to isolate ribosome between E. coli to overcome that's very staph stoichiometric methylation we carried out directed evolution under antibiotic selection to improve methylation of the ribosome and using that we were able to obtain nearly stoichiometrically the ribosome and they're unambiguously based on the crime maps identify position of the newly introduced metal group to our knowledge this gave us a first view of how antibiotic binding pocket is occluded to eight distinct chemically distinct classes of antibiotics through a single methylation. And that knowledge now allows us to identify regions of antibiotics that's terracly clash with CFR modified ribosome, which is valuable to design of next generation antibiotics that are overcoming resistance resistance and with that I'd like to, to thank my lab and especially Vanya Stokovic Keetling Tsai and Kevin McCasker who have contributed to data that I've shown to you our collaborators in structural biology and directed evolution and all of you for your attention so look forward to the discussion after the next talk. Thank you Danica. So I'm going to introduce our next speaker. Fanurios Tamamis, who is an associate professor of chemical engineering at Texas A&M. His research focuses on the computational study of interactions between proteins and modified RNAs and DNAs. So first I would like to thank, thank you for inviting me here and for the chance to present my work and to hear the interesting findings that you have, we have all shared here. So, allow me to start with a brief introduction. We all understand that chemical modifications to RNAs are important in biological and disease-related mechanisms and their critical for health. There are several studies mapping the location and abundance of a kind full of RNA modifications. And particularly RNA protein complexes are central to processes in the cell, and they have been studied experimentally and computationally, and that in modifications which are dynamic and reversible can modulate such interactions and mediator-abid responses to environmental changes. It is thus important to understand the dynamic interplay between RNA modifications, writing, reading, and erasing. So here I'm providing some of the challenges in understanding the impact of RNA modifications, most of which we have discussed about in this meeting. And I am focusing on the fact that it's important to integrate experimental and computational approaches together to help in understanding protein interactions with RNA modifications. So what is the problem we have looked at? My lab, computational in collaboration with Lydia Contreras Experimental Lab. Given a protein, what is the spectrum of RNA modifications that can be recognized, and how they can, how are they recognized? For example, here you can see the target protein which is equal to PNPAs protein. If you see an RNA nucleoside in the question we asked is given that we're looking at this position, what are the different RNA modifications that can be present there and interact favorably with this protein target. So the solution we came up with is this computational protocol characterizing modified RNAs with proteins that operates in two stages. One, we have a fast and efficient screening of RNA modifications binding to the protein in stage two. We have an in-detail examination of selected modifications that come from this screening and using more accurate simulations and financial calculations. One of the inputs we're giving is the structure of the protein RNA complex, the position of positions to modify the library of RNA modifications to examine, and the proportion parameter of RNA modifications. We are identifying high-fin RNA modifications and in the end modeling the complex structure and also being able to derive baphysical insights. Allow me to show some details about some stages of this tool that was published. So here you see the protein and the RNA strand. In the first step I've told you that we had a high throughput like screening tool in which we truncate the protein and we're looking at the small portion of the protein that contains the RNA binding site. We're doing simulations in implicit solvent to save time. We are performing short simulations and we have also modifications arranged in clades. For example, you see the clade of guanine here and you see that the way the tool operates is that if this is, for example, favorable, then we proceed to investigate this. If this ends up not being favorable, then the next modification would not have been explored. Now, in the particular problem we initially entered to solve with Lydia, we looked at PNPAs, this is an RNA strand, and we looked at which modifications can be present at positions 4 and 8. For example, we would modify 4 and 8 simultaneously and this is based on the fact that these are key positions interacting with the particular protein and also based on the baphysical mechanism associated with this protein here. So allow me to go on just briefly to the results. So we had the protocol screening out a big amount of RNA modifications. Then we had selected modifications that we started using detailed simulations, explicit solvent MD simulations and then free energy calculations. And here you see computational results plotted against experimental results. We have a, I would say, a quite good correlation between computations and experiments. And we predicted aid oxygen to be a good binder but also computational results delineated additional RNA modifications that are binding to equal life for example PNPAs. And of course, you can also see that we were able to identify which we were less favorable. So, after solving this problem the next problem we aim to solve with Lydia was the inverse problem, which is given an RNA with an RNA modification. And we designed which we call it can we do computational evolution of the protein so that we can have a protein binding with higher affinity and or specificity to that modification. So, this is technically the inverse problem. And we, we aimed at solving this problem because PNA PNPA plays an important role in supporting cellular tolerance box is a distress. And, however, I wanted to understand that there are key challenges when you're trying to solve such a problem. Of course, you need to, for example, understand that if we are aiming to solve such a problem. Residues in close proximity to the RNA strand. So we might we need to find the, the most critical ones. So this comes we need to select and particular a particular challenge to this system is that we are dealing with a homotrimeric protein and each mutated residue with interact with interact with different parts of the bound DNA. Another key aspect I want to highlight here is that we need an optimistic detail to capture such effects when designing. So long story short, we, we follow the similar approach. Again, we providing a structure fast screening detailed and followed by by detailed simulations and financial calculations investigating now the most promising protein newtons binding to other name modifications and the output is identified high affinity protein newtons modeling and providing physical insights. And, for example, in the end we have proteins recognizing any modifications with improved affinity and or selectivity. So to do this study. First of all, we had to select which are the most critical positions that we would the computationally design and then experimentally test. So initially we used bio physical approaches to select these positions, and we ended up selecting three positions. So we used bio informatics to set to technically to limit the number of modifications that are possible on the protein, and then we went back to bio physics to do screening, and out of the possibilities that came up from by to select the modifications and design the protein to bind and with high affinity to particular modifications and in this case modification was a doxon gene at an investigative at position eight and nine. After sending the results to Lydia we were very happy to see the experimental validation of the results. And we have seen that. So these are our design newtons in comparison to the wild type and we have seen that we have been able to additionally now design a protein with a slightly higher affinity and specific and specificity for a doxon gene with respect to the wild type. And here, you also see our ability after we derive these uses emissions snapshots to to do and to perform a new depth by physical analysis and understand which interactions from the mutations we have introduced contribute most to and has affinity and specificity for the RNA transmission. And here you see a summary of what we found. So these were the key results by Lydia continent's lab, showing that all five minutes actually showed higher survivability to hydrogen peroxide exposure, compared to the complemented wild type PNP as and PNP is variance with enhanced a doxon gene affinity and selectivity, differentially affect cellular tolerance to excessive stress. And this observation provides a clear link between more accurate discrimination of RNA excitation and cell survival to conclude, I would like to briefly say that in this field, it's important to be able to screen for the entire repertoire of RNA modifications. We can be enabled through the topology parameters provided by my girl, as well as sing and ff tool, also provided by my girls lab. We need an efficient and accurate. We need efficient accurate methods to simulate such systems. And we need a domestic detail in my opinion, in this particular problems because we're looking at subtle changes in on the RNA, introducing different groups, method groups, etc. We need to be able to technically understand them, the comparison between one fits all approach to a custom based approach. So in some cases, we may need to be able we may need to, I mean design the system, computationally to be able to to to have the necessary detail we need. So, for example, to carefully decide how we're going to score the energy between a protein and RNA, and we need to, we need to finally understand what is the role of other nucleosides to the RNA modification and this bind to the protein. At the most least, I want to highlight that, and what if, for example, the protein structure is not experimentally resolved. We have new methods, including for example, Alpha fold and Alpha fold database, which now provides an excellent source for us to start at protein RNA complexes. I would like to highlight the importance of Alpha fold would predict predicts a protein structure, yet we need to find and we need to, for example, develop tools, but I mean better tools to, I would say highly accurate, investigate the binding of RNA to protein, and then understand how the modified RNA would bind to this protein. Thank you so much. Okay, thank you both. Let's thank both speakers. We have a few minutes for questions so you can come up or use Slido for those. I will start by asking both speakers maybe to share perspectives that they might have on how this understanding, especially with recognition of RNA modifications might be incorporated into some of the technologies that we've heard about for detection. If you do you want to go through. Yes, okay. Then he said, would you like to go first. Oh, I think you can go go first. So he said, I mean, ideally, I mean, I would see, I mean a potential avenue in the future of us being able to, you know, while understanding how RNA modifications bind to proteins, and while while we're understanding the problem, better in terms of biophysics to be able to, you know, a design products and modify products that can be able to detect the presence of particular RNA modifications. I mean, that can be present. So computationally one, I mean, one could see that, you know, we're just modeling a system. So we're solving what we are, what we're having in the system, but also we're having the capability to change what's on the RNA, what's on the protein and investing in different possibilities. And this could potentially be integrated with other strategies in my opinion. So thanks, Danica. I completely agree with that. I think the specificity of the RNA recognition limits application of some of the tools as broader tools for depositing and identifying sites of methylation, certainly in the case of mechanism based crosslinking very few enzymes go through covalent enzyme substrate intermediate that can be stabilized and, you know, other than the RNA, all its work and fine metal C and our work on our own and CFR this this methodology mechanistically is limited to those cases so thinking more broadly about how we can expand this beyond mechanism directed base is really a great point. Excellent. So we have a question from the room. Yes, this is Glenn Bortrich. So with the PNP aces, this is for fan. I'm a colleague fan. Do have you use that for RIP seek have you have you actually employed it to see if you can pull down. You carry out system. No, we have just used actually we have solved this. We have used this as a model system to be able to start I would say our journey into solving such problems. So, what they should what they showed in the last slide is a problem we're currently working with with Lydia, and his understanding rules of of readers in in recognition, for example, I mean, I don't have in the skin is white age in in in complex with an RNA strand. So now, we're in the process of derive, I would say, having computers deriving RNA modifications which are prone to interact. And then in collaboration with Lydia's lab, I mean experimentally validating predictions. And also, also like to highlight the importance of in such projects is important for experiments. So in other words, it's it's like work. It's important to close the loop. For example, as providing insights to Lydia's lab and Lydia's lab providing insights to our lab so that we can potentially improve the predictive ability of our tools as well. But to answer your question, no, but we have used this system primarily as a model system to start developing the tool for outside our new explorations in the field now, which are mostly focusing on reader and readers and white age proteins. Thanks. Excellent Brenda. This is a question for Donica. And I'm thinking about when you're using cryo em to detect modifications. You know, the auto pickers are going to pick particles and I don't think they're going to be, they're going to be blind to the modification I would suspect. So I'm wondering what level of modification you need in your RNA sample, and that probably depends on resolution to so if you could give me some of those numbers. Yes, that is an excellent question and exactly the reason why we needed to do directed evolution on the hypermethylating enzyme of the ribosome so that we can push from 25% of modifications that while type enzyme has to near stoichiometric modification that we get with with variant evolved in the presence of high amount of antibiotics. So we have not compared variants of different that that metallates a different degree to be able to very quantitatively address your question of, you know, by tracking one modification based on how much it's incorporated by the enzyme. So we did look into wild type E coli ribosome that has been of course very well characterized by other structural methods crystallography and as well as mass spectrometry for stoichiometry or modifications. So we can using cram and QPT max so we can call modifications from well resolved area of the of the map and when they are let's say above 70 or 75% stoichiometry for lower percentage modifications, or in poorly resolved regions of the map and unable to do that. Oh, we got a follow up. I just wanted to say thank you. Excellent. Well, let's thank both of our speakers. And we'll move. Thank you. We're going to move to our next subsection about artificial intelligence. So I'd like to introduce Rafael Townsend, the founder and CEO of atomic AI, which is a company that focuses on the structural elucidation of RNA molecules using artificial intelligence. Thank you so much for the intro, Julius. Let me share my screen now. Just want to confirm that you can see this. It is almost there we go. Great. Fantastic. Well, thank you first of all so much for having me here today. It is my pleasure to speak a little bit about our past work and around AI and geometric deep learning specifically for three dimensional RNA structure. A lot of this presentation is going to be related to my previous PhD work at Stanford that we are now also continuing here at atomic AI. So, you know, very high level just to set the stage. We're talking here about specifically RNA tertiary structure prediction, right, going from a one dimensional sequence to the three dimensional shape adopted by these structures are the ensemble of confirmations and many cases of cancer instead. And really, at the high level, there's a strong belief that 3D RNA structures sort of as highlighted by no fan and others previous talks here that really enable and enhance mainly scientific and commercial use cases in human health agriculture synthetic biology, and beyond. And while there's been some amazing progress made in terms of these experimental structure determination techniques such as X ray crystallography or NMR. They're very expensive in terms of cost and labor, and therefore in practice the vast majority of the structures across another transcriptome and other sources remain in practice inaccessible to us. And therefore sort of a lot of our guiding sort of work here is that the ability to accurately predict these three dimensional structures has emerged as a need. And really following the protein world, and as has been alluded to as well here, there has been some major advances on that front through these alpha fold type technologies that from Google deep mind specifically. And, you know, really what has happened is that there's been these highly accurate AI driven predictions of 3D protein structure specifically, and a key advancement that has really enabled a lot of this was really the reliance amount of gold standard experimentally determined protein structural data, as well as evolutionary type data. And so this is really been seen as a major breakthrough, roughly in 2021. Now, a major issue, when we look at RNA, instead is you know really that there's really the severe lack of known RNA three dimensional structures and therefore standard alpha fold type models have ended up being insufficient for predicting, you know, these kinds of structures, in particular if we do a rough order of magnitude sort of comparison. Now we're talking about on the order of a few hundred or thousands of RNA structures versus the hundreds of thousands of experimentally determined protein structures. And this is really the need in some ways for designing new and specialized computational algorithms to be able to accurately model these systems. So, that is specifically what we did, starting with this work, which we called areas or the atomic rotational aqua variant score, which was really a specialized deep learning algorithm that was that enabled us to yield sort of unprecedented structural accuracy. And specifically some interesting salient points here is that areas itself was trained on 18 RNA structures in total 18 crystal structures specific. And I'll go into the details of that in a second and through this sort of training algorithm, we were able to consistently win these international sort of blind competitions for 3D RNA structure prediction, in particular. To explain the Aries method for a second here, we essentially it is known as a scoring network that essentially consumes a structural model of an RNA, a candidate structural model sort of a hypothesis for what the three dimensional structure would look like. And in particular, it, it looks at the individual atoms that comprise the structure the atoms and three dimensional space essentially. There are multiple layers of deep learning essentially these graph neural network machine learning layers essentially it is able to learn a representation for each atom that encodes its neighboring three dimensional environment. And then from there these features are averaged across the entire structure, further layers of some more standard machine learning algorithms are applied, and we end up with a final predicted deviation from the experimentally determined structure. In this case we measured sort of predicted root mean square deviation from these structures. And what we find and as I'll go into in a second through these kinds of machine learning approaches we are able to outperform or classic physics based approaches at predicting three dimensional RNA structure. So, one thing that we were very interested in measuring here was looking at essentially not just, you know, retrospective sort of accuracy but really looking at the ability of these kinds of deep learning networks to essentially generalize or make you know essentially de novo predictions instead. So, one key aspect that we started testing was this ability to generalize. And I mentioned this before but areas is really trained at the end of the day on 18 small RNA molecules whose structures were published between 1994 and 2006. The average length was about 30 to 40 bases. And instead we bench march areas on much larger RNA molecules that were published between 2010 and 2017 the average length being over 100 bases in this case. And the initial sort of finding that we came with was that areas was able to dramatically outperform, you know, state of the art sporing functions in terms of picking out accurate structural models of RNA. And what I am displaying here specifically is really the comparison of areas as a scoring function to other for physics based scoring functions including the Rosetta scoring function as well as 3D RNA score. And in this case, each of these red crosses represent a specific RNA specific case and we are displaying overall 21 separate RNA is in this case. And is that for any given RNA the top one areas prediction oftentimes falls below five anxious RMSD. And in part, oftentimes actually below two angstroms RMSD. And when we compare that for example to Rosetta, which we can see in the bottom right here, in particular, a number of structures that Rosetta is able to predict at you know a roughly 20 angstrom RMSD or 15 to 20 really areas is able to improve overall. And so so far, all of this to be absolutely clear is purely a scoring approach, right, we are working off of candidate structural models that have been generated by a separate sort of physics based sampling system. Now, the interesting question was whether, you know, if we combine these two and really assess the system as a whole, how that would perform. And so really the next logical step was to enter this, this combined approach into fully blind structure prediction challenges. And so, you know, you may be familiar with cast and RNA toggles being the RNA equivalent of cast that has now been slowly merged into the cast competitions. And, you know, just to describe it briefly when a structure is determined using experimental methods, the results are withheld, and then computational groups are asked to submit their predictions to a third party, which, you know, only after the fact, you know, is then the experiment is released. And therefore we can sort of have this unbiased blind assessment of these computational methods. And we entered areas into four rounds of this competition and found that on all four challenges, you know, areas achieved a higher accuracy than all competitors. You know, just to highlight this again, you know, for one of these given RNAs, in this case, you know, RNA, which was actually puzzle 24 and the Dino virus associated RNA, specifically in this case, we found that areas was able to achieve about a 4.8 angstrom armesty accuracy compared to the next best in this case Rosetta at about 7.7 angstroms. And the other interesting finding that we came from this was that there was no consistent second best method, including sort of structures that were generated by human experts in this case. You know, you know, I always when giving these presentations and told the third showing some more structure, so in the interest of, you know, really diving into that we can actually look at specific detailed tertiary motifs that are recovered by these kinds of methods. And what we can find is that areas is able to recover these tertiary motifs in a way that, you know, was not previously accessible to these kinds of computational methods. And in particular we find that we're able to find motifs such as intercollect T loops base triplets, the loop into helix kind of motifs effectively triple helices, or tightly packed helices overall. And the other interesting piece of this is that we find that areas is also able to spontaneously recover key aspects of RNA structure. In particular, we found that we recovered the optimal distance between RNA strands for optimal base pairing natural, essentially de novo recovering the energy well that matches experimentally determined distances. And we also find that more broadly areas is able to spontaneously recover you know the percentage of Watson quick base pairing in a given structural model, as well as the amount of hydrogen bonding her face. And so, overall, we find that you know, from this area method this areas method we find that this geometric deep learning approach is able to achieve state of the art results and predicting three dimensional structures. We also actually interestingly represents a fairly interesting use case in machine learning as well, where essentially it we were able to train and test these algorithms on a very limited amount of data, as well spontaneously recovering key aspects of RNA structure formation. And while this has been some fairly exciting work for us specifically at this point, one thing that has become clear, moving forward, right is that this problem is far from solved at this stage. We need for additional experimental data at this point to continue training and improving upon these approaches, at the end of the day. And that is really a lot of the work that we are doing here at atomic AI today. Now, before, you know, concluding this whole presentation, let me just give a few brief acknowledgments. These are men there collaborators from the days standard. And, you know, the, especially my advisor Ron drawer and his group, as well as we do DOS, providing a lot of the RNA structure expertise, the Rishi condors lab, and Russ Altman and need on this. And just as a final sort of remark here, just to highlight it for a second really we find that these abundant structures right if we're able to get this ability to predict RNA structures can have really this broad impact across RNA discovery specifically which is a lot of our internal focus here, leaving not only enabling RNA targeting small molecules, but also around the rational design essentially a various RNA based therapeutics in particular. You know, I'm talking about, you know, mRNA vaccines, for example, but also, you know, other systems such as gene therapy vectors circular RNA, and you know, other sort of ASO si RNA type approaches. And I think that's really where I'll leave it for you here today. Thank you very much for your attention. Thank you Rafael. Thank you for our next speaker, which is Mary McMehan, who is a director of biology at Revere therapeutics and early stage biotech startup focusing on RNA and AI technology development to reach disease targets previously considered undruggable. Great. Thank you so much, and I'm delighted to join you today to share some of our efforts in advancing RNA targeting small molecule drug discovery. And I hope you can see my slides here. Yep. Great. Thank you so much. So Revere therapeutics is an early stage startup that is focused on RNA and AI technology development for the treatment of genetically defined disease. Since just starting our lab in Shenzhen in China, we have our AI lab, and most of our research efforts are conducted in South San Francisco where I am located. We have been advancing three main platforms. This is our splicer platform, our binder platform, and our degrader platform. And just to kind of zoom in a little bit more on what we mean by this is we're really focusing on two main modalities to target RNA. One of them is a small molecule that can modulate RNA splicing, and then regulate post transcriptional gene expression. And the second is a targeting structure. So we had a very nice introduction in the last talk about, you know, the different fold and structure that RNA can form. Our strategy is to find the small molecule that combine the specific pocket within RNA and then regulate post transcriptional RNA expression and function. And at the center of our approach is computational biology and AI capability for both of these modalities. And what I'm going to tell you today is that we take a traditional computational approach to guide hypothesis generation for both target site identification and also for building our RNA-focused compound libraries. And again, the AI comes into this in that once we generate data, we can repeat this cycle and we can actually really improve, you know, the target site that can be the best target to drug. And also we open ourselves up to a very big chemical space that can be more, you know, tractable for targeting RNA. And, you know, again, just to kind of come back to the computational framework that we're using for our binder. This is really a 2D structure prediction that is guided by chemical probing such as shape map. And for our splicer, we're really using more of a sequence based approach from in-house perturbation data of splicing to identify the best target site. So one of our approaches is really thinking about the target site for genetically defined disease. And then really levering the AI to build the RNA-focused library. And I think that this is, you know, the area that I see a lot of promise for our drug discovery effort. So I'm not a computational biologist, but in the short time that we have, you know, started the company and screened various targets, you can really see the power of the AI to build the library and to increase the hit rates and validate actually the target site. And for the splicer, this is really coming from rounds of screening data, or sorry for the binder. And for the splicer, this is, you know, virtual screening and computational aided drug design library building. And at the center of our approach, we also have our Voyager server. And this is really a tool that allows us to compute and store the target sites across the whole transcriptome. We also include coding and non-coding RNA and coding and non-coding regions. And we can integrate this into, you know, a genome browser. And this is also curated with public available, you know, human genetic data, but also we have a proprietary set of rare disease or genome sequencing. And the goal here is really to identify and predict for us potential functional sites within the transcriptome that can now be levered for drug discovery. And here is an example of our Voyager server. And what you can see is within any region of the transcriptome, we can identify our predict sites for our binder platform here such as some structure motif. We can also identify target for our splicer platform, including a junction here where you see maybe premature termination codon that would lead to nonsense mediated decay of this transcript. We also annotate, you know, clip data for RNA binding protein interaction. We also can include here mRNA binding sites. And we're also starting to integrate data on RNA modification. So, you know, for the splicer, what we're doing is using junction data sets to really identify the best target site for our candidate. And we really focus on the inclusion of poison exons or exons skipping by alternative splicing. And then using AI deep learning models, we can predict these across transcriptome. We then validate these in cell using functional genomics, and we can conduct cell based screening and compound profiling. For the RNA structure pipeline, we take a different approach. We can take the sequence of the RNA, and then we predict structure. Again, this is mostly 2D based structure prediction. Taking into consideration the local and global thermal stability of the structure. We also consider co variation analysis for structure function relationship. We confirm or redefine, refine the structure using chemical probing methods such as shape map. And then this leads us to a hypothesis generator in combination with functional genomics where we can validate the function of this target site. And now I just want to share with you an example from both our binder and splicer platform, where we validate a target site, and we show an example of heat confirmation. So here's an example of a non coding RNA target that we predicted a very stable structure. This was a multi way junction. And interestingly, a human genetic data told us that this is an important functional region on this non coding RNA, because mutations in that domain are causative for several diseases. We then can confirm the structure using shape map and this 2D structure, and we submitted this region for our bio physical screen. So we perform a number of bio physical screening approaches for our target binder. And once we obtain a hit from the screen, we can confirm that the compound, the small molecule, these are all drug like small molecule that we can identify using shape map, the potential binding site, or the change in structure that is occurring upon compound binding. And here you can see that there's a nice dose response between the alteration in the structure and the compound concentration. We can also confirm using a number of bio physical method, such as ITC and SPR that the compounds can bind the structure. We also confirm with Kim clip in cell interaction of the small molecule and the RNA target. And then here in this case we also see functional activity. So the region that we targeted is important for a ribonuclear protein complex formation and the RNA protein that binds this region is important for stability of the transcript. And here you can see in cell, we see a dose response inhibition of the target so the binder is inducing degradation of the target in cell. Another example from our splicer platform, we identified using our computational approaches, a putative site that when modulated by small molecule should induce the inclusion of a poison exon and activate degradation of the RNA by nonsense mediated decay. We can then create a cell screening tool to to modulate to monitor the activity of this site, and we use this using a splicing light up mini gene reporter. And we can perform screens with 1000 a molecule and identify hit and in this example, we can confirm for a neurological disease target that in cell, the small molecule can modulate splicing of the target site so it includes the small molecule is inducing the inclusion of this exon. And it is leading to decrease in the total RNA for this transcript by nonsense mediated decay degradation. And then we also see a decrease in the target protein expression upon small molecule treatment. So this nicely validate using the binder and splicer our target sites identification, but also our hit confirmation. And what we're very excited for moving forward is utilizing both for our binder and splicer platform AI to basically increase the hit rate, but also to broaden and expand the chemical space that can target RNA. And here is an example of how we have to use this approach. So we can perform our biophysical screen with say a specific RNA structure, we can identify small molecule that are specific and selective for the structure. And you can see here the hit rate is quite low. So this is a diversity library of 14,000 compound that our RNA target is screened against. And the we can then use both the positive molecule and the negative. So the molecule that bounds and the molecule that does not bind. And we can feed this into machine learning model to predict, you know, compounds from broader library that may bind. And for here, for example, you can see some of the input for the model include chemical, you know fragment descriptors. We can then we went to a commercial library of 300,000 compounds, and we predicted based on the training data both the positive and negative compounds that would not bind that we call negative. We selected 1000 of these, and then we selected compounds that we predict should bind. And here we selected three and a half thousand compounds. And what we found was from the compounds we didn't expect to bind. We did our screen with all of these compounds. We find that the negative compounds, again, they did not find in the screen. Whereas the pool of compounds that we predict now might be binder. We actually could find a lot of binders and we found that the generated hit rate almost increase the hit rate by 10 fold. So we're very excited moving forward to applying these models, not only for small molecule that can bind structure, but also for small molecule that can modulate splicing. And just to summarize, you know, this is really helping us advance our drug discovery rock flow in going from our target identification to identifying the small molecule that combined RNA. And the key here for RNA targeting and even thinking about applying to RNA modification is the chemical space that we can use and explore that it will be tractable for RNA. And just to thank you for your attention. And again, the opportunity to share our efforts with you today. And there's a lot of people to thank from our team at our both sites in San Francisco and in Shenzhen. And I look forward to discussion. Thank you, Mary. Let's thank both of our speakers. We're running a little low on time but we have time for a couple of questions and I'm going to start one that summarizes some of the what's in Slido. For Raphael, you know, you mentioned needing more data for AI prediction, we're obviously thinking a lot about RNA modifications and how AI might be used to predict those or make sense of some of the data we're getting. One of your key part of your approach, though, is to come up with candidate models. And we still have a lot of uncertainty in how RNA modifications affect, you know, basic secondary structure predictions, and which might cascade and generate generation of those candidates. So I wonder if you could comment on the needs there. Yes. And I certainly agree there is a strong need for, you know, data around these modifications and their impact on structure. I mean, AI is cool at the end of the day and it's only going to be as good as the data that you feed into it in the first place. You know, to briefly comment on this on the Aries model, you know, if you're atomic we've actually gone beyond this at this point and sort of remove the need for, you know, a separate sampling step essentially and it's much more end to end at this point would be the simple way of putting it going from sequence to an ensemble of confirmations specifically to address issues like that where you're reliant on if you're reliant on a separate sampling step essentially that samplers unable to sample the correct confirmations or really understand chemical modifications. Then that will be sort of insufficient in of itself. Thank you. All right, we have a question from Brenda. I think this will start with Rafael. Mary might have something to say about it but I'm thinking if I understood correctly your training structures were mostly crystallography crystal structures. And so there are, you know, sometimes very non physiological conditions in those crystal structures and I'm wondering if any of your AI methods take into into account, you know, you know, if you if you want to know the structure in the cell is can you take that into account. Yeah, the short answer is yes at this point, you know, I would say that is definitely a major sort of area of active research is the first thing I'll say but you know, being able to model the you know the structure across the cell or contact in the presence of proteins right overall protein RNA complexes and other sort of conditions more generally is really a key need and one that you can build into these algorithms and that we already are. Great, great. Thank you. Excellent question. The question is really similar to the question that Brenda just asked, and that has to do with how do you think about the conformational heterogeneity of RNAs, because like Brenda mentioned a lot of times in a crystal structure we get one structure but I'm thinking of a recent example in the rival switch world right rival switches should be relatively well structured RNAs and you would think these would be great training structures but you know there's a cobalamine in the rival switch that was crystallized. And then, last month, some AFM data came out of NIH that showed that there's actually 10 different confirmations of this rival switch. How do you, when you're training data sets how do you think about the RNA is inherently dynamic. I mean by the way I think what you're doing is great and absolutely important but just in terms of figuring out what the real right answer is how are you going about validating that. Mary do you want to take this one first. Thank you. And so, you know, I think here the key for the structure of course we know there's a lot of confirmation. And from the perspective of I guess the therapeutic would be which one is the most disease relevant. From the basic biology points of view, and you know how do you get to the confirmation and even from 2d, you know, we're using the shape map chemical probing, you see a lot of heterogeneity right even in different cell type in cell. And I think the main question there is the single molecule structure resolution. And, you know, even, you know, advances with nano poor for example this is something that is, you know, we heard yesterday how it's been used for RNA, but also thinking of using nano poor for single resolution of RNA structure at that 2d resolution. But in terms of the 3d, I think the resolution is really important. So, for us, it's more once we have a molecule that we're confident is binding, then we take the 3d approach to confirm structure to get further. But I guess, you know, the more method we can develop to really think about single molecule and resolve those different confirmation. I think that there's work there to be done to move on Nick, do you have a question. I can email them. Yeah, we got to have one quick more question by Nick. Okay, so it's building off of the questions that we've just asked but it just I had it occurred to me that these in the training day that you don't know what you're missing right it's just unknown. However, your models are likely failing because of and I'm not I'm not I'm not saying that they are, if they fail. They're failing because of lack of sufficient training coverage, right. But what that provides potentially and I want to get your like input on this. So if your models failing in a particular area with a particular part of the sequence and you can document that. Does that give us some insights into where we might be looking in collecting new data. In other words, can we use the failures of your AI models to predict where we might need to be looking and doing more research and collecting more data. To answer that shortly. Yes, what's very interesting is these machine learning models can output not only a prediction but also confidence in the prediction in the first place, which is actually quite interesting in of itself. And so there's a sort of framework essentially, and this is a little bit of what we're building here at atomic known as active learning in some ways which is really, you're letting the model sort of uncertainty on certain places guide where you're running the next that you know the experiments are almost like designed to maximally reinforce the AI's predictions but also helps guide you towards where there might be some very interesting novel biology because it's not predictable and was not previously known. So long and winded answer, but you know really. Yes, at the end of the day you can get those you can get those confidence measures and really hone in on those low confidence regions and that's a very interesting approach overall. Let's thank Mary and Raphael again. Okay, our last subsection here focuses on cryo em professor via was not able to be with us today so we have Jeffrey Keeft, whose vice department chair, director of the structural biology and biophysics core facility and a director of the army society at the University of Colorado School of Medicine in May you'll become the executive director of the New York structural biology center. Jeff's research focuses on understanding how RNA structure RNA conformational changes and complex intermolecular interactions combined to enable diverse RNA function. Okay. Great, can everyone hear me. Great. All right, hopefully you can see the slides all right well thank you it's a I guess an honor to be the last speaker, maybe a dubious honor but no seriously it's really great to be invited to be a part of this workshop I've learned a lot it's been really interesting. And as you just heard I'm currently at the University of Colorado, and in May I'm going to be moving to New York City. So, I'll just frame the my presentation by by giving you one slide about what my lab has been interested in and for the last couple decades we're really interested in the role that RNA plays in viral disease. 75% of the viruses that are known to cause human disease have RNA genomes, those that do not have RNA genomes of course are still using RNA within the viral life cycle. Viral RNA is really are littered with interesting sequences or interesting structural elements that control all these different events that have to occur for successful viral infection and so we're really interested in studying these not only to understand how the viruses infect the cell and how they cause disease, but also the structures that RNA is capable of forming and and really how RNA structure produces very specific biological functions so we play with viral RNAs but really I'm interested in RNA structure and how that drives function in general. Now traditionally over the years my lab has used x-ray crystallography as our primary tool of structural determination and we combine that with biochemistry, biology and biophysics. But we really, you know, we've gotten to the point now where we can often fairly reliably solve the structures of RNA, folded RNAs using x-ray crystallography but of course it is a fairly limited technique and that some RNAs are just not crystalizable. Or they're not amenable to the technique and so most more recently my lab has turned to cryoEM. Trying to see if we could use cryoEM as a robust and reliable way to obtain structural information on viral RNAs, especially at a size in which you normally don't think of cryoEM as being a robust or reliable technique. So a little bit about that today, but I do want to of course this is this is a workshop on post-transcriptional modifications. So I'm going to start actually with you know posing up a very maybe simple question, you know, could we use cryoEM, single particle cryoEM to directly visualize post-transcriptional modifications in RNA. And I'll start with a case study which is I guess kind of builds on one of the talks earlier this morning, which is looking at a ribosome. So a non-viral RNA project in my lab through collaboration with the Ryslan lab, we've become interested in the molecular biology of Giardia and as part of that we decided to solve the structure of the Giardia ADS ribosome using cryoEM. A few years ago we put up a preprint where we had a preliminary four angstrom structure and you're welcome to look at that on BioArchive. Not too long after the Yonath lab published a 2.75 angstrom structure, so much better than our four angstrom structure, and now we have a structure right around 2.5 angstroms that we are preparing for publication. And I want to use this as a case study just to show you, well, what we learned from really these three complementary structures of essentially the same target. So, as you can see here, you know, the resolution is really quite good, this heat map of the resolution, you know, we're hovering right around 2.5 angstroms global resolution. But of course in the core of the ribosome if you kind of slice through it here and you look in the core of the 60s, you can see there's parts where we're approaching 1.5 to 2 angstrom resolution. So really a nice high quality structure that we could look at. Now, interestingly, when the Yonath lab published their 2.75 angstrom structure, they proposed sites where they observed post transcriptional modifications and this is interesting because Giardia is a relatively understudied organism. Unlike E. coli or many other organisms for which the ribosome has been studded structurally for a long time and there's a lot of information about it. We don't know, have an issue where there are post transcriptional modifications within the Giardia RNAs or Giardia ribosomal RNAs. But the Yonath lab proposed that they could see these modifications and they provided a table saying this is where we think we see post transcriptional modifications. Now, because we had a higher resolution structure we could go back and we could look at and say does our map show the same thing. And in some cases it did. And in some cases it didn't. So the point to five or so angstrom better resolution that we got allowed us to see that in. In fact, some of these features in the map that looked like methylation sites were in fact, bound water molecules. So we could now distinguish the water molecule from what would be a methyl in other places. It looked good. Our density or our maps I'm sorry, can contain features that were consistent with the modification that they proposed. And other places it was ambiguous and other places we were pretty confident looking at our maps. We didn't see anything there that looked like a modification. So this is I guess, both a cautionary tale and up and a cause for optimism the optimistic side is that yes we could clearly see in our map. Some of the same features they saw for some of these distinctive modifications, especially when the nucleotides that we're looking at, we're in the core of the ribosome where you're talking about 1.5 to 2 angstrom resolution maps. But in other places it was clear that they had been misled. And so that's perhaps a bit of a cautionary tale that, you know 2.75 angstrom sounds good. But in many cases it may not be good enough to see the modifications or confidently assigned. So I think there's hope cause for optimism, or also some cause for for care. I don't show this to to bash the the Yonath lab in any way but just to say this kind of maybe impinges on the question of structure, or I'm sorry standards, and, and best practices, what resolution is good enough. When do we see something in a map that we can confidently say that's a modification. How good does the resolution have to be, how good does the map have to be, and how do we integrate orthogonal techniques. So as a mapping technique of novel modifications cry we am is probably not yet robust in terms of the standards and the practices that are already in place. So let's turn to away from ribosomes because you know we know ribosomes have modifications we're pretty good at studying them know we can get ribosome structures to high resolution. What about other RNAs RNAs that are maybe not complex with proteins RNA similar to the ones that we're talking about that might be embedded in messenger RNAs or might be non coding RNAs that are produced by the cell. In this case, I'll talk about the tetrahymine a thermophilic group one intron. So this is a self cleaving self splicing ribosome, which was discovered decades ago and is heavily studied structurally biochemically, and biophysically. And so we've been able to obtain a 2.44 angstrom map of this RNA. This is in vitro transcribed RNA so it's does not come from the cell. It's not going to have any post transcriptional modifications, but you can see again there are parts of overall the resolution is 2.4 or so with in the the folded core of the ribosome there are regions that are hovering right around two angstroms so very good resolution. And you can see in the maps we're starting to approach the sort of resolution or detail that one might hope would yield visualization of post transcriptional modifications. Probably we're right about at the edge. And again, it's going to be mostly in the core not in these peripheral regions, where the maps are going to be that good so I think this is an important point to bring up is that, you know, different parts of cryom maps are going to have different quality, different resolution. And so one might be able to visualize or assign post transcriptional modifications in some parts of the map. And so one might be able to visualize or assign post transcriptional modifications around the peripheral or the exterior of the molecule or maybe they're being recognized by proteins are involved in interactions, we're going to be much harder to see because those are the most mobile parts where in general we get the lower resolution. But, again, the Tetraheim and a group on intron is somewhat specialized case, it's a very stably folded and well behaved RNA for structural studies. It's something that's maybe a little bit more typical of an RNA that one might want to study. So this test case I'll show you is a tRNA mimicking viral RNA element. It's about 55 kilodaltons so smallish for traditional cryo we am but certainly within the realm of what people are doing now. And I won't go into the virology here but this is the secondary structure. It's found on the three prime end of certain viral genomes and it's involved in various aspects of the viral infection cycle. So we were able to solve this RNA only molecule using cryo we am and this has been published, but the best resolution we could get was somewhere between four to 4.6 angstroms. And that's good enough to see a bumps corresponding to the the phosphates of the backbone. It's good when use good enough when using various density modified and sharpening techniques to start to build models. We were able to build a complete structural model of this 55 kilodalton RNA. But if you look at the maps over here you can see there isn't nearly the amount of detail that one would need to place or to observe post transcriptional modifications. And again, I'll point out this was in vitro transcribed RNA so we don't expect to see any modifications. And there is evidence in the literature that this viral RNA does receive some modifications as perhaps part of its tRNA mimicking behavior. And if that were true we would have been unable to validate that in the in the maps, even if we had had RNA that was actually, you know, authentic from an authentic viral infection. And I'll just say I think this is probably the resolution that that we're really pushing what's what's possible for most RNAs and the other thing I'll also say is that part of the reason that we were limited here was because of what's already been mentioned, dynamic conformational changes within the RNA, and that probably limited our resolution to a great degree. And the last thing I'll say is a technique that we're that we're developing my lab to try and see okay, can we, can we solve RNA structures by cryo and routinely part of the problem is that RNA structures tend to be often the domain that you're interested in studying in a folded domain is fairly small 3020 kilodaltons in size that's still really difficult to get reasonable maps by cryo we am so our idea is to append smaller RNA domains that we're interested in studying on to the group on intron, which I showed you was very amenable to cryo we am. We build these are we in vitro transcribe these chimeric molecules, put them through the process, and hope to be able to recover maps very quickly so the idea is we take the group on intron, we circularly permutate it, we append RNAs of interest and you hope to get a map like this, where you're an RNA of interest is now displayed on this larger RNA scaffold. Long story short, this works. So here's an example of the group on intron with this exonuclease resistant RNA from Zika appended to it, and you can very clearly see in the maps, there's the RNA that we appended. You can also see that the resolution is not spectacular. In fact, we were able to get this to about five angstrom resolution, which was good enough to build an initial model, which fairly recapitulates the crystal structure. So I failed to say that we use this as a test case, because we already knew the structure of this RNA from our crystallography studies. And so we were, we were pleased to see that we could essentially recover a very similar structure using cryo we am to what we had seen by crystallography. So we've applied this now to a few other RNAs, and we can pretty routinely get now maps at about 4.5 to five angstrom resolution, even with RNAs that are smaller than 20 kilo dolphins. So this is not high resolution. It's not truly high throughput. But I'll point out that, you know, we went from essentially making the RNA to having initial maps for these RNAs in less than a week. So one can then one can see the global architecture, perhaps get initial models of RNAs at mid resolution fairly rapidly. Okay, so I'll just finish there because I know we're at the very end of this workshop and I'm sure people are tired. But cryo we am has has clearly can you can see post transcriptional modifications that's not surprising, but it's going to be somewhat slow. It's not going to be routine for every RNA. It's probably limited to RNAs that are in fact well folded the more dynamic the more RNA the more difficult it's going to be. And so far, you know, we've really only done it on in vitro transcribed unmodified RNAs with the exception of, you know, ribosomal RNAs and applied certainly to T RNAs. You know, we need to we need to decide if we're going to really use cryo we have to map or to observe putative modifications. What's the gold standard for resolution, how good does it have to be before it ends up in a database. What modifications look like in a cryo map I was really pleased to see the needs to talk about automated methods for perhaps interpreting cryo we and maps I think that's really important. And then you know what techniques to be trust what orthogonal techniques are going to be required to really make sure that something we see in a cryo we and map is authentic is that mass spec is it recognition by an antibody is it some recognition of those I think orthogonal techniques will always be needed when you're using structural biology methods. And so in the future, you know, maybe we can expand the use of scaffolds or other molecules to display RNAs that might be modified. I think the artificial intelligence and the automated analysis are going to be really important. You know we're going to have to think about if we want to study post transcriptional modifications of RNAs we're going to have to get the RNAs out of the cells where they're produced in vitro transcribed RNAs. So if you're talking about a particular motif in a particular five prime UTR of a particular mRNA that you want to study. How are you going to get enough of that. That is modified at a high enough level that you can put it on a cryo the embryo and actually do structural biology and I think that leads to some question Kristen that's a little bit earlier. And I also think it's going to be important maybe to have more than one structure so you know we were able to compare our structure to the structure from the oneth lab, and that gave additional insight into where modifications were were likely occurring where they were unlikely to occur. And so, you know that gets back to the standards how much do you need to see before you really trust what your, what your eyes are showing you. Okay, I will finish there. And I'll just thank the people in my lab who worked on this particular project to our highlighted yellow, all the support and the funding that we got and I'm eager to see what questions emerge so thank you. Thank you Jeff, let's thank Jeff. So we're close to time but I think I'll ask one put Jeff I appreciate the future perspectives you gave at the end but to follow up on one point this concept of scaffolds and wanting to work on obviously native RNAs. Do you envision a way to combine those approaches to you know capture native RNAs and lock them down so that cryo we can be applied. I'll admit I have not thought about that too much but I think that's an interesting question right so you know perhaps you could get to a point where, if there was a specific RNA structural motif in the cell that you were confident was getting a post transfer from modification you were interested in, and you were confident that the signal that triggers the modification work was contained in that motif. Maybe you could append that motif onto another RNA that's being expressed in the cell, it would get modified and pull it out. But of course there's a lot of ifs there right and so. That's where I think yeah there just needs to be a lot of basic groundwork about what's driving the modification in certain location what's the stoichiometry. Yeah. Excellent. Well in the interest of time and the needs to wrap our workshop up let's thank all of our speakers in this session that was really fantastic. I'm just going to thank all of you that have participated in person and online and all of the speakers. I so appreciate how actively everybody participated. This is hugely helpful to the Academy's committee that has to evaluate this task and I know I've, I've learned a huge amount, and I've also learned what I don't know. So, give yourself a hand please. And then just to thank you so much to the Academy's staff and our AV people back there everything has been really great. So let's