PANEL 3 — Hyperindividualized Treatments | Precision Medicine 2019

Just another WordPress site

PANEL 3 — Hyperindividualized Treatments | Precision Medicine 2019

ISAAC KOHANE: All right, everybody Let’s settle down here And I’m about to hand over the reins to Alexander Haig– I mean our– SPEAKER: Are you going to be here all week? ISAAC KOHANE: What? SPEAKER: You going to do some shtick? ISAAC KOHANE: I’m not going to do any shtick You’re safe I am very pleased to introduce the person who’s going to relieve me of all responsibilities for the rest of this session Our leader here is a graduate of HST MD-PhD, worked with our stem cell biologists at Harvard, and then went on to be involved with several companies in the genomic space and tech space And most recently, he has become senior partner at Takeda Ventures, and also is father of three lovely girls And any mistakes he makes tonight are purely the result of his ketotic diet Thank you very much [LAUGHTER] DAVID SHAYWITZ: Success with that diet, man All right Looking forward to this So just before I start, this is my favorite conference of the year I was trying to think about why that is And I think it’s the most haimish conference I attend, which may say a lot about some of the others But I think it’s really the combination of incredibly high quality science, but really built around an incredibly deep humanistic essence, and a core, and fundamentally around the patients And it obviously reflects the characteristics, vision, and warmth of the leadership, and of Zak, in particular So if we could just sort of start by giving some recognition to Zak and the organization of this conference [APPLAUSE] All right, so this panel, it’s– we’re going to get to it because it’s unbelievable talent– is about hyperpersonalization, or hyperpersonalized therapeutics I just kind of– one slide We kind start on the one hand, it’s the sort of the vision of future work into– I don’t even know what that is– but that in the future, everything is going to be these hyperpersonalized treatments You show up and some 3D printer in the back is going to manufacture exactly the right drug for you at the right time In contrast, I thought it was worthwhile to compare this perspective from Harvard geneticist Richard Lewontin, which he wrote in the New York Review of Books in 1992, when he was sort of– it’s a very famous essay called I Dream of Genome, which was essentially about the– and in his words, the very first thing he starts off with is the definition of a fetish– and he was basically making the point that scientists were fetishizing the genome, which now has arguably been extended to fetishization of data tech and all of this stuff And he essentially is saying that there is a tendency of people to over-promise what the science can deliver So in that context to lead us through this, we have an exceptional panel, where first Tim is going to talk about what’s possible George is going to talk about what are the key barriers Ken will provide the population perspective My Bay Area friend Matt is going to provide the entrepreneur perspective And then Kim is going to provide what is arguably the most important aspect here, the real discussion of the patient impact So with no further ado, Tim TIM YU: Hi, folks It’s a real pleasure to be asked to participate in this august panel I get to open us up By way of introduction, I’m a neurologist, a sometimes biometrician, and a geneticist at Boston Children’s Hospital I run a research lab there And it’s my privilege to share with you a case that I’m really excited about It’s been in the news a little bit recently So if you’ve read about it, please bear with me But hopefully, we’ll get to discuss some of the details But I think the reason I’m excited about sharing this case is because it speaks to a time in the field where our disparate practices, genetic discovery, clinical care, advanced informatics, are coming together in a potentially very exciting way that is really allowing us to actually say, this word “precision medicine” may actually mean something So without any further ado, let me introduce the case and then move along So this story begins in January, 2017, on a Friday evening I finished a day of work, and my wife turned to me at the dinner table and said, would you take a look at this post? This was a post that was put up on Facebook by a Boston Children’s emergency room physician, Jessica Flynn

Deede And she had actually passed along a message from her college classmate, who described a family’s really terrible plight in Colorado And this is the story of a family with two beautiful children The eldest, six years old, Mila, had just been diagnosed with Batten disease, a neurodegenerative disorder without a cure And they had found one of the mutations in the causative gene But they needed to identify the second defect for the definitive diagnosis, and also to test their four-year-old son, Azlan And they were looking for a lab that could help them do whole-genome sequencing And this is something that we had been doing on a research basis for quite some time And so I looked at the posts and I clicked on the link And I looked at the website that the family, following likely Matt Might’s instructions, had set up to share their story, and learned about her mysterious decline This was a young girl who was very, very healthy and energetic and active as a young girl, who then began developing slowly, insidiously, vision problems, gait problems, and then at age six, accelerating symptoms with worsening gait following speech disruptions, slurring of her words And then using fewer and fewer words And then stopping using her words, behavioral regression And then actually, within the span of a few months, she actually went blind I learned from that family website that her clinicians in Colorado had given her a partial diagnosis They had done clinical genetic testing and they had found a mutation, one mutation in this gene called CLN7, which was a known disease-causing mutation for a disease called Batten disease What is Batten disease? Batten disease is a rare subtype of a rare family of disorders They are recessive genetic disorders that are inborn, and they are caused by mutations in one of 14 different genes that control the ability of our neurons to control waste products, to recycle waste And when that system goes awry, unfortunately, patients with that disease develop vision loss, seizures, motor and cognitive decline, beginning around age 3 and 1/2 And then unfortunately, predictably, with death by age 11 And there were no available treatments, because there were really only about 70 known cases of this disease ever published in the literature So our lab saw this post and said, we want to try to help And so that led us down a remarkable journey that I’ll present to you in very abbreviated fashion In essence, in the next few months, we performed whole genome sequencing out of our lab at Boston Children’s And we found that the normal CLN7 gene locus, here, it did carry the one mutation her doctors had already found But I mentioned that this is a recessive disease, and I failed to explain that recessive diseases require two mutations to manifest So the question was, where was that missing second mutation? And it turned out that whole genome sequencing revealed that there had been a second hidden mutation found in her noncoding DNA And this noncoding mutation here, which is inaccessible to clinical sequencing as performed today, turned out to actually be the key second hit that caused her disease It disrupted the gene splicing, or the assembly of the instruction set required to make this entire 13-block gene, essentially truncating it after block 6 and leaving her with no residual CLN7 protein So now as a geneticist and a genomicist, a lot of the times we would stop here We’d found a precision diagnosis And we would report this back to the family and say, we’ve solved your question But the unusual opportunity that we realized was afforded to us was to ask ourselves, can we actually fix this? And the concept that we pursued was that her genetic instruction set was entirely intact, except for the presence of this mutation here which caused the cell to ignore the entire downstream half of the gene And we began reading intently about a technology called antisense oligonucleotides [ASOs], which had just shown wonderful effects for another neurodegenerative condition called spinal muscular atrophy And antisense oligos are short, 20-base synthetic RNAs that can home in on parts of the genome And they can essentially do various things, including knocking genes down But they can also change splicing And the very simple concept that dawned on us we ought to try to do, was to develop antisense oligos that would actually bind to this abnormal site here

that was introduced by her mutation, block it, and thereby allow the cell to now continue to read through the rest of the gene Compressing a lot of work into one slide, we found that we were able to do this We met our patient in January, 2017 We identified the mutation about a month and a half later We obtained patient cell lines a month after that And by a month beyond that, we had proven that there was this splice gene assembly defect caused by her mutation And we began thinking hard about the antisense oligo approach By September, we designed and identified ASOs that actually were working to block abnormal splicing in her cells in a dish And these were working with a potency similar to that described for other antisense oligos that had been FDA approved And then we were able to validate that this actually not only fixed the RNA-level defect, but also made her cells healthier With a lot of assistance from a lot of folks in the pharmaceutical industry, who now emerged from the woodworks to help us with this case for this child, we developed a plan to do rat toxicology manufacturing and FDA filing in a matter of four months And at the very end of January, 2018 we began dosing our patient A couple of comments We named it, of course, milasen, in honor of our patient And this is one of the very first examples of a truly personalized medicine It is, as far as we know, the first drug created and approved for a single patient purely on the basis of her genome sequence I think that, with assistance, we put together the fastest manufacturing campaign ever for a clinical drug And we used a rolling FDA submission process, whereby we manufactured the drug, began testing it in animals, and before we were done with the testing in animals, we had started injecting our patient because there was so little time to lose And this whole process has seemed to have paid off We are now about a year and a half into the trial And we’ve seen signs of disease stabilization and seizure improvement through the course of this year and a half And I’m just showing you one of these primary data plots, showing that at the onset of the trial, our poor patient was having up to 30 seizures per day– between 20 and 30 per day And they were typically lasting two minutes at a time 30 of those per day Each of these red stars represents escalating doses of a drug that we gave to her over the initial dose escalation loading period of our trial And as this went up, we saw the frequency trend down And in addition to the frequency, we saw the intensity lighten as well And by June and July of the first half year, she was having between 0 and 10 seizures per day And the seizures were not lasting two minutes They were lasting 3 to 10 seconds each This effect has been stable now over the last year of treatment We’re still adjusting the dosing a little bit to get the frequency a little bit lower But it’s been early signs, but very, very positive signs that this drug is working So that’s the provocative case This is an individualized drug made in one year from a molecular diagnosis, and then all the way through to an individualized medicine that we were given permission to proceed with by the FDA It defined along the way a scientific, clinical, and regulatory path for treating N of 1 cases This hadn’t truly existed before We have paths for repurposing drugs, but we don’t have paths for approving agents on this time frame when they don’t yet exist And it allows us to even begin to speculate about visionary ideas, clinical genetics, interventional genomics of the future, whereby this type of procedure, this type of drug, could be applied and developed to other children with orphan diseases, especially those that fall under the umbrella, under the threshold of what is considered commercially viable This is something that we’ve managed to do as an academic lab And is there a potential to continue doing this in a way that allows us to sidestep some of the challenges in delivering this via traditional industry? And it does raise this theme which was mentioned earlier today Could drug development actually have gotten to the point where some pieces of it, some pieces of the tool set could be robust enough that we can actually engage them almost under the practice of medicine, not under traditional models

So I’m going to stop there I think I’ll leave lots of ethical and legal and social questions for discussion among the audience and our panelists But set the table [APPLAUSE] DAVID SHAYWITZ: Thanks, Tim That’s absolutely extraordinary stuff And speaking of absolutely extraordinary stuff, our next speaker is George Church, who will talk about some of the key considerations and barriers in this space GEORGE CHURCH: All right I don’t have any slides so I’ll just sit for now And I would say, we are facing in precision medicine the most expensive drugs in history And we just heard of a $2.1 million per dose drug And I am assured, and I believe, that these are cost effective relative to the previous way that these diseases were treated So gene therapy is cost effective relative to other orphan drugs and all of the other care giving costs that are– and this is being economically studied That doesn’t mean, just because it is cost effective relative to standard of care at the present, doesn’t mean that we’re done My group has spent most of my life trying to bring down the price of many parts of biotechnology For example, DNA sequencing is now 10 million times cheaper than it used to be And it was also about 10 million times better quality So at the beginning of the Genome Project, we were aiming for a 1% error rate Now we’re closer to 10 to the minus 9 So this can happen Well, what are the ways it can happen for these hyperindividualized treatments? Well, hyperindividualized, if you go that far, then you maybe have something that is the practice of medicine and not only doesn’t require a full FDA trial, cannot be done by an FDA trial, because there is only an N of 1 So that’s one way to bring down the cost The second way to bring down the costs is by eliminating therapy completely So you can, for some of these diseases, they can be anticipated You can do genetic counseling So this is the second part of the revolution here We have all kinds of ways of writing our genome, or our -omes, to– in a therapeutic sense, we can also read them and avoid them at earlier and earlier stages To some extent, another example of hyperindividualized treatments are cancer, neoantigens You could be the only person on the planet with that particular antigen that your CAR-T cells or other immunotherapies can attack And that also could be practice of medicines as N of 1 But to some extent, it sometimes is too late By the time you can detect cancer DNA in your bloodstream, or by the time you can sequence enough to get a neoantigen, or by the time you have a child whose brain has been compromised, it’s kind of too late to have a therapy And your odds are slightly against you So if you move earlier up, then you might be doing newborn screening, or fetal diagnosis, or pre-implantation But even earlier, it can be before you get married But that’s crushing news to find out the love of your life is not compatible with you– 5% chance It’s about a 5% rate of rare diseases, rare severe Mendelian diseases That means that 5% of fiances will get very bad news You can move even further up to create a list of people with whom you’re compatible Again, we can have an ethical discussion later on But the point is, earlier is better for preventative medicine And these can be hyperindividualized and very inexpensive in the sense that the cost of sequencing is close to zero now, close enough that society can pay for it It’s hyperindividualized in the sense that– and even FDA approved– in that the costs of false positives for preconception, premarital, predating, is close to zero If I make a mistake where two people are incompatible,

just add it to the other 3 billion people that I’m not going to date So I mean, it sounds funny But this is oddly powerful medicine that is oddly not really part of the FDA So I should answer the question of the whole meeting, which is, can I accelerate precision medicine? And to show that I have a counterpoint to my own point, which is, yes We’re using it routinely to develop better gene therapy vectors, like we’ve made 1.2 million AAV, adeno associated viruses, which is one of the main therapy delivery We’ve made them custom by using AI machine learning And it’s working spectacularly I’m actually embarrassed how many years I spent on energy minimisation methods, which are now completely swept away by these methods But anyway, so I can, at the same time, advocate gene therapy and avoiding gene therapy DAVID SHAYWITZ: I look forward to your talk next year on “gdate.” MATT DE SILVA: Thanks so much, David So I feel very strange standing up here My name is Matt, and my background is in behavioral economics I took AP bio, so it got about as far– I connected with Matt’s story this morning So I was out in California because I took a job with Peter Thiel, working at his hedge fund, and was going on doing that for a couple of years after graduating, basically trading currencies, until my dad was diagnosed with glioblastoma So what brought me to this podium today– and I really will try to also not have just the tech founder in the hoodie, which I’m not wearing, sorry– perspective, but also the patient’s perspective Because Notable Labs, the company that I started, was really a last ditch effort to save my dad And so the very brief version of that founding story is that when he was diagnosed, he had three separate brain tumors And for those of you who have dealt with trying to enroll a loved one or yourself into a clinical trial, they have very strict inclusion and exclusion criteria And unfortunately, multi-focal disease is an exclusionary criteria for most of the drugs that were being developed at the time that my dad was diagnosed So immunotherapy was out Any kind of targeted therapies for GBM were out What was left were all of these drugs that were being repurposed, and we’ve heard a little bit about that today An example of this was a drug for pinworm that was being developed at Hopkins for brain cancer patients that had responses There are lots of other drugs around the world So about 50 of those is the list that I ended up with And I went to his clinician in Rochester, and I said, OK, there’s 50 drugs being repurposed for glioblastoma as single agents And my parents don’t want to travel to get those drugs So pick three And I’m like, let’s combine them together I was then, of course, met with like a, who the hell are you? You know, this is not how we practice medicine, which makes total sense And so I went looking for a way to rationally combine those 50 drugs And so what I ended up backing into is this field that’s been around for a very long time It traditionally has been called chemosensitivity So we put some drugs in some cancer cells And then we try to predict who’s going to respond, and who won’t It has not worked very well in the past And when I went out there looking at the companies that I could use for my dad’s case, all of them had one-size-fits-all solutions So it didn’t matter whether it was breast cancer, or brain cancer, or pancreatic, or whatever Yet, the advancements, specifically in brain cancer biology and translational models, had gone far beyond what was available commercially So here at Dana Farber, or at UCSF, there were some really interesting models being employed, but they were not available commercially So that was really the impetus behind starting the company, was to hire a post-doc Start a lab across the street Have a surgery for my dad Get the cells out Do the drug testing Treat the patient It worked in terms of, we created the models, we found drug combinations Unfortunately, the turnaround time that it took us to do that was time that my dad didn’t have And so we lost him very early on, and all of the other brain cancer patients that we tried this in So we needed to move to a cancer type that was going to be addressable much, much more quickly And so we are now focused entirely as a company in blood cancers So our general process is that we’re taking peripheral blood or bone marrow from these patients We are screening thousands of different drug combinations in a fully automated laboratory I appreciate all the conversations today about scalability We rank order the drugs that are active based upon specificity And so what I mean there is that,

do the drugs target the patient’s leukemia cells relative to their healthy white blood cells? And then we do all of that in a time that is actionable So we’re targeting always doing things within a week Since this is a panel on hyperindividualized treatment, I wanted to talk about a case, which is a public case This was presented last year at a pediatric cancer conference called ASCO A patient, a 16-year-old with acute myeloid leukemia that had a bone marrow transplant from his sister, and unfortunately relapsed six months after that transplant, which is a very poor prognosis He was being put on a standard treatment as maintenance And while he was on that therapy, the clinician sent us his sample for drug testing And we found that a combination of two drugs which are approved for multiple myeloma, a different type of blood cancer, were active in the sample And so when the patient progressed on that standard treatment, they decided to implement the multiple myeloma drugs And the only reason they were able to do this is that, fortunately, that combination had been tested in kids before And so this is a big barrier towards hyperindividualized treatments is, do we expect that we need to have seen that in a clinical trial before we treat patients? In this case, that was able to happen, and two weeks after starting that therapy, achieved a complete remission that was MRD negative And it was very durable, which is why the case ended up getting presented Moving from N of 1 to N of more than 1 is really important with all of this, for us in oncology And so we’ve been now doing trials that test whole cohorts of patients So we’re just kind of actually coming back from Amsterdam, where this was presented at the European Hematology Annual Congress We’ve just completed a trial with Stanford showing 83% and 85% positive and negative predictive value in an adult blood cancer called myelodysplastic syndrome I’m trying to now scale this up to see, OK, does it not only predict responses, but if we use assay guided therapy, will patients live longer than with standard of care? So I’ll stop there, and look forward to the discussion on the panel [APPLAUSE] DAVID SHAYWITZ: And now to talk about the interface between– there’s all of this discussion and the promise, the articles, the coverage of Tim and George and everyone that we hear about And then someone comes in with, oh, I have this diagnosis, or I have this concern How do you sort of reconcile the sort of the promise and then the individual fears? And to discuss some of that, Kim KIMBERLY LEBLANC: Yes, thank you, David Hi, everyone I’m Kim LeBlanc I’m the associate director of research operations at the Undiagnosed Diseases Network Coordinating Center, which is here at Harvard Medical School I’m excited to be able to share a bit about our experience working with patients with very rare conditions in the Undiagnosed Diseases Network, and as David said, what the impact on them will be for these hyperindividualized treatments For those of you who may not be as familiar with the Undiagnosed Diseases Network, or UDN for short, it is a multi-site research project that’s funded by the National Institutes of Health We have multiple different clinical sites, a sequencing core, and basic research groups, that are spread out across the country, all focused on diagnosing patients who have been to numerous medical centers without finding answers In the UDN, the diagnoses that we’re making are often very rare Sometimes there are more common conditions But we have also diagnosed and discovered conditions that are new to medicine For the conditions that are established, there may be known standards of care, recommendation guidelines for treatment And in those cases, the clinical sites will recommend those treatments But that, unfortunately, isn’t the case for many of the patients who come through the UDN So for those cases, what are we doing? What can we do? One of the first steps for these cases is connecting them with support groups that may be out there They may be established support groups They may just be Facebook pages for a gene of interest Oftentimes, these groups, these families, are experts in these rare conditions, and can work with clinicians and researchers to collect clinical information that will be necessary for research or therapeutic opportunities down the road In the UDN, we also, if you’re not aware, use social media to try to find other patients with some of these very rare conditions, some of the conditions that are new to medicine, to try to put them on that first step of finding others For these rare conditions, it also takes bridging the clinical and research divide in many cases To mention, how understanding the mechanism of Mila’s condition was helpful in discovering a potential therapeutic And for many of the patients that we see in the UDN,

it’s the same case We do have a model organism screening center in the network that’s able to investigate the underlying mechanism of disease in some cases We also try to partner with researchers outside of the network who may be experts in the condition or disease pathway in order to learn more about the condition and what treatments may be effective Matt also mentioned the work that he’s doing at University of Alabama And he is one of the principal investigators of the coordinating center So something we’re trying to do now is scale up the approach that he’s using at UAB to benefit more patients in the UDN And my last slide– importantly, patients and their families are crucial partners in this work This is a picture of our Participant Engagement and Empowerment Resource, our patient advisory group in the UDN in my opinion, we have to partner with patients and families at every step in the process, from the very early research stages to when a therapeutic becomes a possibility These people are the experts in their conditions, and can collect the important clinical information that’s necessary to move all of this forward So thank you And I’m looking forward to an interesting discussion [APPLAUSE] DAVID SHAYWITZ: So I’m going to start off with a couple of questions And then I’m going to turn it over to the audience, because there have been so many good questions throughout the day at the other panels as well And I guess maybe the first, sort of for Tim and George, is, how should we think about reductionism in biology? I feel like we’ve been moving from a sense where, oh, we’re going to get a gene, and the gene is going to unlock the secret to everything And now we’ve been going to a more sort of what seems like a nuanced and sophisticated view, that it’s not just the gene But now it almost seems almost like a Back to the Future, where now here’s this example where you– here was a mutation, you really fixed or made a lot of progress into curing someone with a very specific gene We’re seeing other examples, with some Mendelian disorders, you know, with hemophilia, with others How generalizable do you think this type of thinking is for patients with disease? GEORGE CHURCH: Well, I think it’s healthy for the field to think about things as being complex But it’s unhealthy to ignore the simple solutions as they present themselves So it might be, instead of one gene, it could be two It could be three Yes, there may be environmental factors There may be thousands of genes involved But very often, the lesson of pharmacology and even genetics, is sometimes, a small number of gene products will be enough So for example, height has huge environmental and genetic components But nevertheless, there are seven diseases that are treated in the clinic with one gene product We also are entering a phase where the multigenics, even if it doesn’t have us a one or two or three gene solution, isn’t necessarily that hard Because we can control the environment And we can now add it up to 26,000 sites So that’s our latest record for editing of the human genome That doesn’t mean that we’re in complete control of those 26,000 sites, or any of the off targets But it definitely makes you rethink what’s feasible in terms of multigenics So at least testing hypotheses is suddenly in a new realm You know, I think that’s what I would say about how we should be thinking about this multigenics and multi-environmental components DAVID SHAYWITZ: So Tim, how generalizable do you think this is? I mean, was it one of those cases where just everything went right? Because it seems there are so many implicit factors that were right Or could this be, with a little bit of work, scaled so that many more therapies are delivered? You know, how do you manage expectations? What are appropriate expectations for physicians, patients, and researchers in the context of the work you presented? TIM YU: So I think it can’t be emphasized enough how complicated this project was, and how complicated these efforts are Clinically, scientifically, manufacturing, from a regulatory standpoint, there are many points of failure And we were extremely lucky to have things lined up in a way that allowed this first test case, this first proof of concept, to go through But I do think that– and getting back a little bit to the question of simple and complex genetics, and is everything going to be– at the root of it, this was a simple single-gene disorder

And that’s what made our managing of the other human and logistical and systems factors, made it feasible If it were something more complicated– digenic, trigenic, more complex inheritance– yeah, obviously, we wouldn’t be doing this But I think that examples like this do provide us an opportunity to develop tools and platforms and gain experience with the interventions that we use Antisense oligonucleotides are at the moment now undergoing really an explosion of interest I mean, this is a field that has been in existence for 30 years, but only now in the last two years, have the fruits really begun to ripen DAVID SHAYWITZ: And it was supported with research funding? Because I’m leading to the next question I can imagine getting a letter from (COUGHING) UHG or somebody saying that, thank you very much for your thoughtful proposal And unfortunately, this is an investigational treatment, which we are unable to support How do you view this as a payer, Ken, or at least representing a population health perspective? When do you see this is ready for prime time and reimbursement? Or like, how do you think– not the answer to this case, obviously And you’re clearly pleased for the patient But what’s your framework that you think about this type of innovation? KEN EHLERT: Well, it seems like your question is not one of what does the funding model look like during our research phase? But rather, when you have something that works and is known, that it’s time for a scale, how do you actually move it to that phase? DAVID SHAYWITZ: Well, how do you make that decision when it’s the science is a work in progress? KEN EHLERT: Yeah, that’s a really good question So there’s a number of different things around it The insurance world is actually– there’s some complexities in there that are not well understood But there are some decisions that people make up front to say, what is it that’s actually going to be covered in the insurance world? If I was to make it about car insurance for a second so that everybody can think about it rationally– [LAUGHTER] You guys have no idea how dangerous it is for me to be sitting up here AUDIENCE: We appreciate it KEN EHLERT: Yeah, Zak is very persuasive When you go and buy your insurance on your car And it’s a lovely 1992 burnt orange Toyota Corolla that’s awesome, and it works for you around Cambridge And you get into an accident And you call up and you say, hey, I got into an accident I’d like to get my car repaired They value the car They write you the check And you move on It doesn’t even occur to us that well, actually, I don’t want to drive a 1992 Corolla any more I’d like to drive a brand new 2019 Lexus instead And so I’d like to propose that, since I was a very unique case, that I should actually get this Well, you could actually write an insurance contract that would allow you to do exactly that That’s not a hard financial thing to figure out how to do it The question would be, is what would happen to all the 1992 Corollas on the road? I would submit that probably most of them would run into stop signs or something along the line In health insurance, you’ve got some similar decisions that actually have to be made So decisions around what to cover, those don’t get made on a daily basis Those get made– they bulk up And then they are processed in batch to determine, should we actually be covering whatever the condition is or whatever the treatment is that actually comes with it? But to this particular case, when you’re thinking about this thing– or a case like this– you’re thinking, OK, so what is enough? If there’s evidence that you have something that actually works, then there’s a conversation that can actually take place about, should this be in the commercial contract or not? Until the evidence is there, putting it into the commercial contract would be inappropriate for those that actually purchase that commercial contract But once you know that something works, now that discussion can actually start in earnest And we can say, OK, this is an interesting therapy And then all the other machinations around it will start What’s an appropriate price, those types of things DAVID SHAYWITZ: When you hear about, I mean, really authentically, when you hear about something like this, as a person you’re incredibly excited for the patient But is your sort of almost, I guess, professional view, wow, this will be an incredible thing for the health of so many people, potentially, and reduce all the costs associated with the burden of care? Or is it more they’re concerned that it’ll kind of sort of maybe work, and you’re going to be winding up just with huge increased cost? How do you sort of see the burden of this from the perspective of a payer?

KEN EHLERT: Sure Yeah, the personal one is obvious It’s awesome I’m a dad I’m a grandfather I think it’s really cool From a professional perspective, what goes through my head is, to build on something George was talking about earlier, how far back do you go? Professionally, I think about what we do as fundamentally, it should be in the disease interception business And the earlier in that process that we can actually implement good diagnostics and great therapies, the better off society will be Now we have some decisions that we have to make around that And those decisions are really hard decisions Decisions like, you know, how much are we willing to pay for something like that? So take the new SMA therapy, at $2 million It is a miracle drug And to say we can take somebody who would normally die at 1, and maybe they can have a productive life, 50 or 60 years For $2 million, that seems like a steal of a deal to me And I would sign up for that for my own children for somebody else’s Maybe I would feel differently if it was my 80-year-old father, who is at the other end, and it’s only an extension of life by six weeks or something along those lines But there’s questions around how to do that That part’s not about is it moral or right or those types of things The question comes down to how When you have a risk contract that’s built around a 12-month cycle, that’s really hard to convince somebody that economically that they can make an investment of $2 million and never be able to actually recoup that That has nothing to do with for profit or not for profit That has to do with, from a capital deployment perspective, if it’s not there, you run out of money eventually So we can treat the first five people, but nobody else But then there’s other questions around how should a therapy like that actually get priced for the market And what’s the appropriate thing to go through? I think Tim’s done a fantastic job in this case of demonstrating new ways to do things Those are those are really interesting questions But at the end of it all, you have to sort of boil those things down and say, OK, we see something It works We’d like work to work for all And then to put the math into real perspective for people, if we made it financial, if we started to look at a therapy like a $2 million therapy for– what is it, probably 350, 360 kids a year? So somewhere in there So we are into $700 million to $800 million to treat everybody that’s born this year with that condition That in the scheme of the overall system doesn’t seem that extreme If you divide that across every man, woman, and child in America, it’s about $2.00 for each one of us How many people here would willingly take $2.00 out of your wallet right now to save everybody that has SMA born this year? Exactly How many of you would take $2.00 out of your wallet right now to save everybody that has a rare condition in every single thing? How much would that be? AUDIENCE: A lot of money KEN EHLERT: 7,000-ish rare conditions And so we’re at about $14,000 We think health care is expensive today And that’s a societal question Should we actually go that far? I have a personal answer My professional answer would be, it’s a societal question and let’s vote Now, I don’t mean politically vote I mean, vote from a bunch of consumers in a room together How do we want to do that? DAVID SHAYWITZ: Thanks I really appreciate you taking the time to answer these difficult, thoughtful questions in such a considered way, Ken I know it was a little bit on the spot I appreciate it KEN EHLERT: That’s all right DAVID SHAYWITZ: So Matt, I have a question for you You sort of started your business– it’s sort of an entrepreneur business question, really You started your business as kind of a hyperpersonalized diagnostics service model almost, right? But as it’s evolved, it’s becoming more of a– as I understand– evolving towards kind of more identifying novel therapeutics Is that right from what you described? What were the considerations? And were there thoughts about not pursuing sort of hyperindividualized therapeutics as a model? Or were there other factors that led to the way your very successful and rapidly growing businesses is evolving? MATT DE SILVA: Absolutely So I would characterize that decision as an “and,” not as an “or.” So we continue to believe in the hyperindividualized N of 1 So can we identify a drug or a combination for that patient, whether it’s experimental drugs or off the shelf? The driver for going to larger numbers of patients is because it’s much easier to tune

the parameters of your assay So making a prediction on whether or not the patient’s going to respond to that therapy or not is very, very hard to do on an N of 1 basis You don’t have a control So how do you really know if the optimization that you’re doing in the lab is making your product better or worse? And when we look at the history in our space, again, it’s just all these one-size-fits-all solutions It’s also not just across different– I described different types of cancer as being one size fits all But even within a cancer, like one specific type of leukemia, we work in AML mostly Eight drugs approved in the last year Some work via differentiation Some are antibody-based therapies, immune therapies, all kinds of different things Having one test that measures the activity of all those drugs is extremely difficult because their mechanisms of action are all quite unique And so by tuning on a drug by drug basis, we believe the accuracy is going to be much, much stronger than if we tried to take that one-size-fits-all approach And so that ends up meaning that if we predict the outcome of a clinical trial for our pharma partner, or for an academic who’s running a study, well then maybe we can solve this problem of not having enough patients for clinical trials, especially with all of the different combinations that are being developed today in oncology Because like for example, with those eight drugs being approved, what is standard of care in AML is actually now an open question DAVID SHAYWITZ: I have a question for Kim And while I’m asking you the question, maybe people could line up if they have questions for some of the other panelists So my question for you is, as you’re talking to patients who are reading about all of these advances, what has been your experience with the expectations that they– I mean, I feel like there has been a series of issues like this– first, the Genome Project, and then your SDI-570, that’s precision oncology, and then IO, and all of these different things where you hear about this great promise that, on average, isn’t going to work for most individuals What has it been like kind of at the point of impact? KIMBERLY LEBLANC: A lot of it is talking with patients and families, and seeing where they’re at and what their expectations are And not necessarily tempering their expectations, but presenting what the opportunities out there are There’s so much more available now than there was even five years ago And I think Matt’s work is a great example of this, where his team will take any patient and discuss what the correct plan may be And sometimes that, too, is describing more of what is underlying that diagnosis, what the mechanism is, and why certain approaches may or may not work Putting that in terms that are understandable for everyone DAVID SHAYWITZ: Do you feel like people are showing up with any type of grounded expectation? Or do you think they’re sort of thinking like, sort of like TV-level miracles? KIMBERLY LEBLANC: We work with a unique group of patients I think in your standard clinical genetics clinic, it may be a lot different But in the Undiagnosed Diseases Network, people have been through such a diagnostic odyssey already that many don’t necessarily come in even expecting to get a diagnosis, let alone a treatment So I think we’re working with a unique group of patients there DAVID SHAYWITZ: That’s your perspective? KIMBERLY LEBLANC: Yes DAVID SHAYWITZ: Zak? ISAAC KOHANE: So I know it’s very tough to give patients bad news, that things don’t work And a question here for Matthew first, which is, how do you feel about someone coming to you with a tumor, and you say, nothing is really going to work? Because that’s going to be true a lot of the times And I have to say, insurance companies have paid for therapies that we all knew were not going to work, like bone marrow transplantation for breast cancer, which ended up costing a lot of money, torturing a lot of women, and not really changing outcomes How do you think about that, about being aggressive, doing your best, while not hawking false hope? MATT DE SILVA: Yeah, that’s actually our current obsession inside the company So I’m really glad that you asked about it So the term here is negative predictive value If it doesn’t work in the lab, it doesn’t work in the patient And so the concept is, yeah, you test all these drugs and maybe nothing looks good That does happen Has happened And that is challenging news for that oncologist and that patient Of course, we’re not directly treating the patients So those conversations happen outside our walls But I think that, from a patient’s perspective– if this was again, my dad, or my daughter or my son there was nothing that was available through the drug

screen, well then at least I wouldn’t be using a drug that– every drug has toxicity Every drug has cost for that patient And we could be going to look through other methods in terms of clinical trials or using genomics or things like that But I think fundamentally, actually, the space that we’re working in is going to be better at negative predictive value than positive predictive value If it doesn’t work in the dish, doesn’t work in the patient is much easier than works in the dish, works in the patient So a perfect example of this is, when my dad was diagnosed with glioblastoma, he was given standard of care because, which there is a biomarker for that predicts non-response quite well There’s only about a 10% benefit rate to this biomarker called MGMT with the standard chemotherapy and radiation Why aren’t we doing upfront clinical trials in those patients to give them other agents, right? And so that’s an example of a negative predictive value biomarker The paper came out approximately 10 years before my dad’s diagnosis Yet he was still offered that same standard of care rather than onto something else So I think actually– DAVID SHAYWITZ: So sort of in a way, you’re saying– MATT DE SILVA: That needs to shift And that we give different drugs So it’s negative predictive value to standard of care, but hopefully positive predictive value to something else DAVID SHAYWITZ: So I think, so your question was, isn’t it kind of gruesome to tell people they don’t have any hope, and how confident are you in that? But then your perspective is, well, if something really isn’t going to work, better to know that upfront so that you can advance potentially to something more effective or something at least different Work to provide sort of better guided advice MATT DE SILVA: And we have to bring these kinds of approaches to the frontline setting That’s where the patients are going to be much, much more likely to benefit from a precision-based treatment than after they’ve gone through three lines of therapy DAVID SHAYWITZ: Thank you for that Adrian? AUDIENCE: I have an easy question What should the role of the FDA be in hyperpersonalization? And does AI change that or not? KEN EHLERT: Go ahead, Tim Are you going to bite? TIM YU: I’ll take a first stab at that So I can speak to our interactions with the FDA around the particular platform that we deployed for our patient And the FDA has a number of roles Clearly, they exist to protect the public health And when you put that in contrast to the title of the panel, Hyperpersonalized, Hyperindividualized Treatments, then there’s a little bit of a disconnect So there’s less of a public health impact for some of the treatments that we’re discussing Now whether you consider, however, the treatment is the drug, the molecule that we injected into our patient, or the process, however If we now are talking about the process of building a drug for a patient with that disease, customized to her mutation, now we’re talking about a place where the FDA clearly has purview And so I think they serve a valuable function and valuable oversight over this type of work I’ll just throw out one other scientific piece, which is not lost on us The FDA also sits in a privileged position They are a trusted third party that has access to data from many different pharmaceutical efforts by many different self-motivated companies And they protect that They don’t share that freely with the public But they sit on that data and can guide other companies along the way without violating intellectual property, towards hopefully a better goal, more efficient goal And I think that’s another critical role that they play in individualized medicine DAVID SHAYWITZ: And just to give a quick additional perspective– and I want to get to the next question– but I think an interesting perspective is that, for years there was this perception that the FDA is sort of really resistant to innovation, or is going to just block and be super-conservative And, you know, over the last several commissioners, in multiple administrations, in both parties, I feel like there’s been this profound switch And there’s a huge interest in pulling innovation responsively, as they always emphasize, but really trying to make sure they’re pulling things forward And to the extent that when I give talks sort of like in different pharma groups, I actually think a bigger blocker is some of the sort of conservative mindset within some of the pharma regulatory groups, internal the pharma, as an example, than as an agency that’s really trying to look at new things Go ahead AUDIENCE: Thank you for your wonderful talks So my question is, the science and the [INAUDIBLE]—- DAVID SHAYWITZ: Could you introduce yourself? I’m sorry To continue Zak’s tradition AUDIENCE: I’m studying the genomics of aging Besides the rare disease and cancer, how about hyperindividualized longevity projects,

and aging-related disease treatment? Will this open a new channel and opportunity for hyperindividualized [? procedure ?] medicine? DAVID SHAYWITZ: So I guess the question is, how does hyperindividualized medicine relate to longevity, and which youngster’s blood do you want injected? [LAUGHTER] GEORGE CHURCH: So yeah, I’ve done a little bit on this I think one of the possibilities for longevity is that it is a core commonality that we have in common In fact, all of us who don’t die of– people in industrialized nations, about 90% of us, will die of diseases that don’t affect 20-year-olds So it could be some common set of mechanisms Its probably not simple, but it could be common In which case, you don’t have hyperindividualized You’re at the other end of the spectrum, where you have the first true blockbuster drug in history, which all 7.7 billion of us could take and drive down the cost That said, the way it’s configured right now, the way we treat diseases of aging is extremely fragmented and individualized We might have a little of each I’m not sure you want to drill down much more than that, other than to provide the counterpoint that it might actually be hypoindividualized DAVID SHAYWITZ: Thank you, George Quick, introduce yourself as well SAM OMARI: Hey, my name is Sam Omari, and I’m an investment banker from Connecticut And we just started working with health AI companies And one of the things that we’re very interested in is the fact that there are a lot of people that will be very old in the future So that’s going to be a big strain on the hospital systems here in America and across the world So where do you guys see AI solving that issue or problem? Thank you [INAUDIBLE] KEN EHLERT: I’m sure I’ll help [LAUGHTER] DAVID SHAYWITZ: Maybe George is going to say– any comments? All right, so we, [INAUDIBLE] KEN EHLERT: Actually, so I want to take that one and go back to the FDA question, too Because actually, I think you have the role of the FDA And I think there’s some things that can actually change that would be really helpful So for example, when we do something today, we publish a trial And you have a treated group and a control group And it works on the measure enough that we can roll this out to the public, and it’s safe enough, and all those types of things One of the things I think that would actually help that would be a little bit of a hybrid is– and I’m not sure who should actually go into and do it– but actually figuring out exactly who it worked for Should we be sequencing people? Should we be digging down into which proteins were expressed? And what is the biomarker that actually would indicate that this thing actually does work? Because when you look at some therapies, you have, it works for 25% of the people, but it works really well And then you have other people, it doesn’t work at all And some of them are cheap Some of them are really expensive The amount of resource it would create to be able to go in and start using those trials to actually dig in and say, this is who it worked for And then maybe start figuring out why I think that’s a weakness of how the system operates today DAVID SHAYWITZ: I think there’s a huge amount of interest in trying to identify the actual outcomes, and trying to figure out the relevant correlates with the outcomes, and trying to figure out who are those populations of patients I actually worry a little bit that people are essentially, on the basis of relatively few numbers of outcomes, they’re looking at an infinite number of variables, over-fitting the data, and then making a sort of a little bit of naive predictions So I think sharing that aspiration, but trying to do it well ISABELLE CHAMBERS: I’m Isabelle Chambers I’m a high school student interning at DURABIO And my question was kind of about manufacturing with personalized medicine And I know, doctor, you talked about this a little bit But do you think in the future that there will be maybe small Simos in hospitals? And how do you think the manufacturing of the drugs and biologics will evolve? And do you think AI will have any impact on that? DAVID SHAYWITZ: That’s a great question Thank you for asking TIM YU: That’s an excellent question And I think that it’s really exciting to think about the technologies that could be deployed at this bespoke local microbrewery kind of way And there are a couple of technologies that have historically navigated that I would say that– well, I’m sorry Antibody therapy, not at the bespoke level yet

Now there aren’t manufacturers building antibodies within hospitals for patients But the level of expertise that we’ve developed with being able to develop the monoclonal antibody against whatever protein target we’re interested in, and then deploy that in a safe way, that capability has really expanded in the last 10 years or so And it took a long time for that to develop I think that there are other platform programmable-like medicines like antibodies that can be used in multiple different ways Antisense oligos are just another example of that And with time, the more we invest in those types of reusable drug platforms, the more we’ll be able to do exactly what you said DAVID SHAYWITZ: I mean, ultimately becoming not quite a commodity, but something that’s more closer to that than being exotic Next question DAVID PLATT: Yes, my name is David Platt I teach computer science over there My question is mostly for George, and a little bit for Ken And that is, thinking of your ability now to look at the genome and predict the possibility of these unusual diseases, especially when combined with a particular other person, one that requires two recessives, it seems to me that you’re now well within the range of something that would actually work in modern society and pay huge, huge benefits We already get tested for– I myself get tested for Tay-Sachs before my own wedding It is not at all an unusual kind of thing And if they’re going to stick a needle in my arm once, they might as well look for the rest of it while they’re there The popularity of 23andMe, for example And I can just see this thing becoming routine Ken, how much would your insurance company pay somebody when they turned 18 to go get yourself sequenced so you know who you are compatible with, and who you were not? Are we not in the hands of the marketeers now? Can you see offering two months free when you sign up if you agree to get your genetic profile? And when you wink at somebody, they’d say, sorry, no, that person is not compatible I can very easily see this slotting into society as just something that is done routinely Do you see things evolving that way? GEORGE CHURCH: Do you want to go first, or do you want me to? KEN EHLERT: Yeah, sure DAVID PLATT: It’s a natural outgrowth of the things we do already DAVID SHAYWITZ: Ken, I think the question is, are you guys going to cover counsel now? KEN EHLERT: I think my part was how much would I pay for that? DAVID PLATT: Yeah, OK How much would it be cost effective? KEN EHLERT: Am I the one getting married? DAVID PLATT: Oh you were going to pay for the 18-year-old to get this test so that– KEN EHLERT: You’re the guy that wants the Lexus, aren’t you? DAVID PLATT: Yes KEN EHLERT: OK, if anybody sees a burnt orange Corolla, it’s his Actually, that one comes down to how much would I pay, or how much would an organization pay, as a risk holder, to actually say can you avoid some of these types of things? DAVID PLATT: Precisely How much would it be worth to you to be cost effective? KEN EHLERT: If you had a one-year risk contract, the answer is, you would pay nothing Now I’ll help you out on that one If I have two people that are going to be getting hitched in some way or another, and presumably they’re getting attached because they’re thinking about producing offspring DAVID PLATT: Yes KEN EHLERT: And presumably, they haven’t produced that offspring yet And let’s just say it takes two, three, four months for that to come about, and now somebody is pregnant One thing that humans have in common is nine months of gestation So that’s kind of cool I just moved you out to 12 months from the time you took that test So actually, if I have a one-year risk contract, I’m paying for something that actually has no benefit to whoever holds that risk contract Does that makes sense? So a way of actually saying that is, what would that be worth if somebody had a risk contract that lasted for 10 years? That one’s actually relatively easy math If you had somebody for 10 years, you would ask, what’s the prevalence of those types of things in general? You would ask, what’s the cost of actually dealing with those types of things? And boil that back down and say, you know what? It probably is actually worth it to pay some amount of money I don’t know what the exact math would be, but you’d be able to actually calculate what that would actually look like DAVID PLATT: There was a guy here last year, maybe the year before, who was saying exactly that You need to have futures contracts on your own health to make that kind of thing cost effective KEN EHLERT: Yeah, it’s a different type of insurance contract than what we carry today in health Actually though, it’s a fascinating concept Because you think about childhood disorders, and it’s not just a feel good thing From a society perspective, taking care of children is, I mean, that’s a foundational bedrock thing from a society perspective There are probably models that we need to explore that says, OK, we have somebody that has a rare condition We’re going to send them into an undiagnosed disease network Who is going to hold the risk long enough to make sure that this thing can actually work?

And I think there are plenty of models that could actually be applied People ask questions about AI and blockchain, those types of things Can you put together a risk contract that allows you to say, you know, this transfers? And there’s a transfer payment that keeps coming back And that becomes a funding mechanism for dealing with people I think those things are actually feasible DAVID SHAYWITZ: I mean, there’s a flip side to this, too, where there is a very interesting JAMA paper that Imran Haque and colleagues published when he was at council, where he talked about saying, OK, well you know things like screening for CF, carrier screening is covered apparently But screening for rare diseases, whereas if you integrate the long tail of rare diseases like has been discussed, it’s less rare than the chance of having a child with CF But in general, that isn’t covered So he was sort of making the argument for that type of consideration I see we’re running out of time I want to make sure we have time for our next speaker DAVID PLATT: Thank for your candor and your look behind the curtain DAVID SHAYWITZ: And thanking the panel Yeah, incredible panel [APPLAUSE]