This week I am talking to James Lu, M.D. Ph.D., President, co-founder of Helix (@my_Helix), an end-to-end population genomics platform. Dr Lu has an interesting background with a PhD in genomics working with Robert Gibbs from Baylor College of Medicine and worked in the Machine learning laboratory at Duke. Even in the early days of sequencing it was challenging to capture data and integrate it into the healthcare system to make it useful and this challenge formed part of the early ideas for Helix
We delve into gene sequencing, the fact this data set does not change over time (and we talk about how it might as well) and detail why gees are not determinist but rather influencers of disease since most genes are probabilistic versus binary in effect
As genome sequencing became more widespread and economical our gene reading capabilities improved dramatically and Helix not only manages the data generation but importantly decouples this from the data usage. Getting to the first and only FDA Authorization for a Whole Exome Sequencing Platform which was driven in part by this decoupling and allows for additional use cases over time of the data as we learn and understand more.
Learn how they achieved their progress by building proof points including this recently published paper focusing on 3 diseases: Population genetic screening efficiently identifies carriers of autosomal dominant diseases, understanding what is impactful and can have a positive influence on patient outcomes
Listen in to hear their incremental steps to create a digital resource library and manage physician workflow effectively and how they have been working their systems 24/7 in support of the COVID19 effort, identifying 50 cases of variants in the USA and correctly predicting the rise of Variants of Concern (VoC) in particular the B117 we are seeing now across the country.
Listen live at 4:00 AM, 12:00 Noon or 8:00 PM ET, Monday through Friday for the next week at HealthcareNOW Radio. After that, you can listen on demand (See podcast information below.) Join the conversation on Twitter at #TheIncrementalist.
Listen along on HealthcareNowRadio or on SoundCloud
Raw Transcript
Nick van Terheyden
And today I’m delighted to be joined by Dr. James Lu. He is the president and co founder of helix. James, thanks for joining me today.
James Lu
I appreciate it next. Thanks for having me. It’s really exciting to be here on incremental healthcare.
Nick van Terheyden
So, as I do with all my guests, I think it’s important to understand people’s background, like so many folks that I have the privilege of interviewing, you’ve got a fascinating background, you’re doubly qualified. Tell us a little bit about your journey and how you got here. To sort of share your experiences.
James Lu
I’m happy to, you know, I would love to say that it was pre planned, but certainly was not I think it only makes sense in retrospectively but, you know, prior to helping to start the company idea Linux, and that was started in 2015. I was I just joined the faculty at Duke University. And I was starting a laboratory focused on machine learning methods are now called kind of AI type methods on on medical records, as well as translational genomics. And I was supposed to finish my clinical residency and my clinical training, they’re really focused on internal medicine, and then ultimately into kind of care and pulmonary. I think I guess I’m proud to say I’m a residency dropout. So I’m no longer qualified actually, to see patients. But I think the medical training and obviously, the research components have been a big contributor to what we’re doing here at helix. Prior to that, trained as an MD, PhD, most of my PhD work was mostly on what was then the very early days of genomics. So I’ve ever worked with Dr. Richard Gibbs, who’s my primary advisor, and he was he runs the human genome sequencing Center at Baylor College of Medicine. And, you know, that was one of the few places in United States doing very large scale sequencing back when sequencing was really brand new. And so that laboratory sequenced Jim Watson, I think, at the time, he was the second person in the world being sequenced, and I worked on some of the early projects of the first, you know, 10s, and hundreds of people being sequenced. And a long time ago, I spent a little time in venture capital, a little bit of time in pharmaceuticals. And a long, long time ago, I was trained as a chemical engineer at Stanford. So that was the origination of my career, but kind of had this interesting trajectory kind of throughout very focused on large scale computational ideas, how do we leverage data to make it useful? And how do we think about application of new technologies to healthcare?
Nick van Terheyden
So, you know, there’s a bunch of things in there that are really interesting, not least of all the sort of early phase when sequencing although, you know, dates back to the first individual, and, you know, the big genome project, and I forget what the numbers are. But, you know, if it’s not billions, it’s certainly a lot to sort of sequence that one individual. And then we’ve seen this, you know, incredible decline, as you were sequencing the hundreds 1000s of patients, had the costs fallen sufficiently, or was it still at the point where it hadn’t really dropped to sort of expand the access?
James Lu
Yeah, it was kind of in this intermediate point. So just as a reference point, you know, back when the human genome sequencing project was done and completed, and the reference was done, I think in 2003, I think that project costs something on the order of $3 billion. And yet today, we we sequence routinely, I think, around $600, for a genome, at least on the research basis. And so, if you look at the curve, it looks like Moore’s on Moore’s law. And it just for the audience, here, Moore’s laws, the doubling of transistors per chip every 18 to 24 months. And so you have this halving of cost effectively, for performance. We had a, you know, sequencing had a curve that was exponential on top of an exponential. And so the cost had dropped dramatically. But I remember, you know, back in the day, we’re doing the first pediatric Rare Disease exomes. You know, you’re begging and borrowing just to sequence the parents, because that really helps you educate the case, understand the pediatric case, but, you know, sequencing and additional individual costs, you know, 10s of 1000s of dollars at that point, so not not billions of dollars, but 10s of 1000s. And so, you know, it was very hard to do now, we obviously routinely sequencing the 1000s of people per day, I’m sorry, not 10s of 1000s across the country. So we’ve come a long way. But I do think from a cost component, just one piece of that, I think that the knowledge component increases exponentially as well as we reduce costs. And I think that’s been a big driver for increasing utility there.
Nick van Terheyden
So we see this, you know, exponential decline, you know, I think the projections based on that curve is that it will very soon be about the price of a toilet flush, which opens up new opportunities in that space. Tell us a little bit about helix and the formation of that company. What was the genesis Is that and how did you arrive? At the idea?
James Lu
Yeah, I mean, the ideas really started percolating in the early 2010s. And, you know, at the time the world was sequencing, what was the first kind of 1000 people. And I think the central insight really was that, you know, in the world of healthcare, there probably is only one data set, that doesn’t change with time. And that actually is your germline genome. And, and that’s the genome that you inherit, and, you know, uniquely, if you can generate the data, a very high quality, and then leverage it in the healthcare system on a digital basis, you can come back to that over and over again throughout your life. And they can form different parts of your care as you kind of progress. And so, you know, helix is formulated on that idea. Well, we call it kind of call sequence, what’s query off it. But there are two main components to what we offer. The first one is, can we generate high enough quality data to make sure that we can use data over and over again. And the second part is what does the data platforms look like to enable the health care system to use that data over and over again. And I would say that the first part of that data generation is a problem that’s gotten easier with time, particularly as technology has gotten better. But the part about orchestrating use of data is something that I think continues to be a major challenge, certainly for people who are working within the American healthcare system. And so we spent a lot of time focused on that data platform and utilization of that data platform.
Nick van Terheyden
So there’s a number of things sort of buried in there. And the first one that sort of jumps out at me, and I think it’s important to explore for those listening is, you know, you talk about the data set, that’s not going to change, you know, a couple of things in there. One, do you think that’s ever going to be different, you know, obviously, with gene editing, and, you know, put the ethics to one side, because I know, huge challenges, but I mean, I think conceptually, it’s possible. And then the other is, if that’s an unchanging model, I think most people, those that are old enough, will, you know, snap back to Gatica and think, Well, wait a second, are you saying I’m pre programmed to end up at this? You know, there is no change? Because this is my genome makeup? And, you know, I think there’s good reasons not to believe that. But I think it’s important to explore and tell us a little bit about your thinking around that.
James Lu
Yeah, happy to I mean, I maybe I just take, make a very clear stance, which is that the genetic code is not deterministic. And so this, this idea that you’re pre programmed, I think is, you know, for lack of a better word is false. Certainly, there are pieces of the genome, which we know on disease do influence the likelihood, and in some cases, very strong likelihood of disease manifestation. But most things in the genome are probabilistic. And they interact with the environment. And they helps, you know, to some extent, I think people describe this setting the stage, and all your interactions with the rest of the world and all these other what we consider other omit technologies, that do have a time component of them all come into play, and it influence both the day to day experience, as well as the the either diseases or the behaviors that you exhibit over time. And so we certainly view the genetic component as something that will be informative for various pieces of your care over time, but we don’t we definitely don’t think that for most of us that this is a something that’s set in stone when you’re born.
Nick van Terheyden
So I mean, I think entirely correct. I think people sort of struggle with that, because, you know, there’s some element Well, that’s my programming, and, you know, you don’t end up but, you know, back to the sort of point of, you know, changing the code. What about diseases? I mean, we have diseases that, you know, we can point to a single point of genetic change, that you could now you’ve got to get it early enough. So there’s all sorts of chat, and I know, I’m way out there. But, you know, it’s not often I get people with this level of insight. You think there’s potential for that to come about putting the ethics to one side and the challenge that surrounds it?
James Lu
Yeah, I mean, I would say just on gene editing, and maybe I can give a little context, right, what we’ve gotten very good at is obviously reading the genome. And the second P is like writing the genome, which I would consider gene editing. And that technology has progressed very rapidly in the last couple of years, starting with something called CRISPR cast nine, right. And I think there’s, you know, Jennifer Duda. And that team kind of won the, I think, within no Nobel Prize last year, I think it was last year on this one. The technology there is incredible, and there is the potential that you could be thinking about ways to edit the genome and ways to eliminate these very point mutations that result in very deleterious diseases. But I do think we have to be very careful about where you apply that technology and make sure that Technology is applied to the right set of problems. We’re often I say, you know, well, I think we all think is dreamed about, which is like, oh, we’re gonna wrap, you know, we’re gonna edit all sorts of places and you know, do all sorts of stuff, kind of more of a hypothetical science fiction type basis. I, it feels a little bit like technology searching for a problem versus the problem. I’m getting so often. So what I don’t want to do, and I don’t think that anyone in the field wants to do is introduce secondary issues. Right, as well as the technology is not perfect, right, the fidelity is not 100%. And so you do introduce other errors, potentially, through the writing process. And so I don’t think that’s something that we understand quite well enough that we would think about it from a general basis. But there are I think there are very specific problems around some of the germline diseases that can be worth having this discussion around.
Nick van Terheyden
Fantastic. I mean, I think, you know, exciting, but obviously, we have to sort of think very carefully about how we progress, you know, lots of challenges. And I don’t want to sort of diminish any of that. And I’m not sort of advocating for one way or another to be clear. As I say, not often I get people that really insights to be able to ask those questions. And I know, you know, certainly one that comes up a lot in the conversation. So you’re back to helix. And you know, what the purpose is of that organization, you sort of teased a little bit about the challenge of gathering that data, the challenge of sort of making it useful, the fidelity or the accuracy. Tell us a little bit about the platform and what you’ve done and what you’ve managed to achieve so far.
James Lu
Yeah, happy to do I think, I think one, maybe I’ll talk about the two components. One is the data generation side, and one is the kind of data platform side, and how they kind of intermingle and some of the things we’ve achieved there. I think one is sort of the data generation side, we’ve been very focused on making sure the data can be used over and over again, to generate the platform for that. And so he went to things that we were quite proud of earlier this year is to achieve the first FDA clearance for for an excellent platform. And I think that was fairly unique. And that if you read the documents carefully what they’ve done, and essentially, a few things that we’ve actually advocated for the first one is, the FDA has accepted the concept that you can decouple generation from usage. So you can generate a data set in the past that high enough quality and then you can use it for an application site in the future that may not have been contemplated the time you generate the data. So that the coupling, I think is fairly unique in the molecular space, because most of the time you generate data set using immediately, and you don’t come back to it. So I think that’s first, the second one is this idea of digital use, right? So we’re going to read query the data, develop a new application or current application, launch them as we go and make any kind of a digital resource and building a regulatory pathway around that one. And so I think those concepts conceptually and then together, really enable world we’re thinking about the usage of those data over time for each individual and generate a regulatory quality that you can just set for a wide range of regulated applications in healthcare. On the usage side, beyond the regulatory side, a lot of it is around how do you manage the physician workflow, particularly as we imagine this technology be used by frontline positions? So genomics I think, rightfully like most technology starts in the specialty space, especially stages, you know, a bunch of power users who have specific workflows and smaller patient populations. And I think we’ve achieved incredible things in oncology space, some of the rare disease spaces, pediatrics, etc. Now, when you move out to primary care, and you using it as a primary genetic screen, or to identify people are at risk for disease, your frontline physician now needs different sets of tools to be successful. And so we’re very focused on how do you kind of work on an interaction? And how do you educate physicians and patients to do that? How do you make sure it frankly, you get the outcomes that you want? Because none of this is a new in many ways. And I think this is a problem that not only will impact your ex, but also do you think about other molecular other omics technologies or other data technologies coming down the road? How do you build a bridge between the frontline physician, the patient and the healthcare system, to enable them to adopt new technologies that we know will improve patient care, but doing a way that’s efficient doesn’t take 30 years and new medical training? And so our ability to develop new technology and deploy is or certainly bill ecology is certainly much faster than our ability to deploy and adopt. And so how do we kind of accelerate the adoption curve I think, is something we’re very focused on.
Nick van Terheyden
So for those of you just joining, I’m Dr. Nick the incrementalist and today I’m talking to Dr. James Lew, he is the president and co founder of helix. We were just talking about the platform, the exciting FDA clearance. I mean, that’s not an easy journey. There’s certainly been some stumbles and falls along the way. I’m sure that didn’t help you. Or maybe it did. Maybe there was some insights from that. I’m curious to understand was the some elements in that past that you look back and think well, from Have an insight perspective to get it through that process and start to deliver value. Were there specific things that you brought to the table? Or that you changed or adapted to that really changed the way the trajectory for helix? And, you know, obviously, this FDA clearance? Yeah, I mean, I
James Lu
think the, you know, much of the things that we’ve been saying for this idea around digitally digitizing the genome, I think, has been an idea that we started propagating in 2015. I think it’s taken, you know, frankly, it’s just taken time to educate the various constituencies and stakeholders around, how do you even think about the digital resource over time. And I think to some extent, the macro trend has been that healthcare is becoming increasingly digitized those those, what many people call the exhaust of the healthcare system. So the claims data, that laboratory data, etc, are actually resources that could be used to improve the care of patients and, you know, either in the day to day services component of it, or even for things like drug discovery. And so we’ve been writing a little bit on that macro trend, but I think we’ve been very focused on the genomic aspect that specifically because we do think that one is unique in that aspect. And so part of it, the education of it, part of it is, frankly, building the proof points. And so, you know, for example, one of the big things that we’ve been working on demonstrating is that as we go wide, certainly for everyday access, one of the key components and key questions has been, does it make sense to screen every person in the United States genomically, and for what conditions and what things are impactful? And so we had this very nice paper that we put out in nature medicine, I think it was mid last year, actually the middle COVID that we wrote with whenever a big health system partners renowned health on the population, about 30,000 people and healthy in northern Nevada, where we screened every single person, unbiased, who walked into the program looked at the incidence of genetic disease, across just three conditions, we looked at whether or not they would have been ascertained in the normal health care system, and what the disease outcomes look like for these patients. And just to ask the very fundamental questions of does this impact clinical outcomes. And what we found is something like 175 individuals are carriers of a very deleterious genetic marker for one of these three conditions. And the conditions are Lynch syndrome, which is a syndrome of cancer, hereditary breast and ovarian cancer caused by braca, one, bracket two, and camileo hypercholesterolemia, which is inherited high cholesterol. So just, you know, nine genes, three conditions, one and 75 people carry, we thought about 80%, or 90%, would not have been ascertained in the medical system at all. So, they either did not have family history, or were not asked family history or would not have been discovered until disease manifestation. But we know that these patients can be monitored and intervene early on to prevent. And so I think it was a pretty big finding that one was, you know, greater than, you know, it’s like, one 1.33%, but that that 90% 80 90% level tells us that the healthcare system is not equipped to detect these individuals. So genomic screening becomes basically probably the one and only tool to really identify early risk. And Shouldn’t we be doing something to identify, and I think in the system we use in the US, typically, patient populations that are highly morbid, that can’t be detected with technology, we should go find those patients to do something about them. And so even though they’re on a narrow basis, our belief now is that there’s very clear benefits from a clinical perspective, we think that’s clear economic benefits for the healthcare systems writ large. And this is something we should be deploying. And so we’re seeing it actually, in all of our programs now, you know, renowned, renowned, is doing it universally, Mayo Clinic, at least in our program with tapestry 100,000 patients is also doing for the call a CDC tier one. So this set of diseases screening universally for that program, we’re starting to move that to other health systems. Now, I think we’re at the very early stage of saying, Look, it’s going to be a general standard of care benefit over time. And then we’ll be able to leverage that data over and over again, in a wide variety of different cases.
Nick van Terheyden
So important to understand, and I will put a link in the show notes to the paper. But what’s the sensitivity and specificity around those? You know, you talked about the nine genes and the three diseases? Well, and specifically, I’m focusing on the false positives, does that even occur? I mean, is that a problem? How did that show up in the study? Yeah, so
James Lu
the sensitivity, like if he, if you look at all the potential carriers of those genetic carriers of markers, I think there were probably around 90 to 95% sensitive, I think there was probably a small set. If you expand the gene list a little bit or expand the way you look at it, you can probably expand slightly, the specificity which is I think, a really Interesting point here. And it’s super important because you don’t want to ascertain a bunch of people who actually won’t manifest disease. Right. So we have very stringent criteria about which genetic mutations or variants are likely to be protein disruptive, and that would have manifested disease. So we think the specificity is quite high. And when we look at the disease population, so we just took an all comers you look at all all people who had carried and you look at just, you know, incidence of disease, we found is something like 26 27% of them had already manifest disease, which is, you know, 10 to 10 to 1510 times relative to the normal population distribution. And so if you just took that snapshot in time, you actually do find that he has incredible amount of disease burden in that population. And the question then, for the rest of the population is would they manifest with more time? And and what we see is, yeah, we strongly believe that would happen. And so when you draw these curves, it looks like something like 50 to 60 70% of people, as they get older will manifest some level of disease, because genetic.
Nick van Terheyden
Interesting. So I, you know, huge opportunity. And I think, you know, that list is obviously going to expand as you start more, in the few minutes remaining. Tell us a little bit about COVID-19. Because obviously, you know, there’s a element there. And that’s, you know, ever present. What’s been going on what’s been your experience?
James Lu
Yeah, happy to I mean, in the last year, helix has become one of the largest diagnostic labs for COVID-19. But I think most recently, actually, we’ve been a big, big partner and contributor with Illumina and CDC, on surveilling the country. And I think one of the one of the key things we’ve been very focused on is variants of concern for early detection, frankly, for not only this pandemic, obviously, but infrastructure for the next pandemic. So we’re building some of those data and sequencing capabilities, right up our alley sequencing side, on making sure that we are ready for that, not only for the continuation, and hopefully end of this pandemic, but also for anything in the future. And so we did one of the key organization identification of the the, what’s the 117, which is, you know, probably, unfortunately named the UK variant.
Right.
James Lu
And, you know, certainly we made predictions in I think, February, this timeframe, and it was just published so that this will be dominant in the US, by around early April. And that’s what we’re finding now. That, unfortunately, this recent uptake, and in in cases here in the US is likely due to both more contagious strains coupled with reducing restrictions on on, on, on individuals from a social distancing perspective, as well as increased travel. And so, you know, for if anything, right, increasing surveillance allows us to get a picture of whether or not we’re going to see frankly, continuation or, or longer, longer tail, or essentially, what’s happening here in the US. So we continue to test regularly. We’re doing, you know, we’re offering 24, seven a day on a diagnostic side. And certainly on surveillance side, our goal is to continue to sample as deeply as possible to really detect any of these variants of concern.
Nick van Terheyden
Wow, that’s, you know, I think hindsight is always 2020. But, you know, not surprising when, you know, we look look at the data today to sort of think about b 117. And the increased transmissibility. And even, you know, there’s there’s increases in morbidity that we’ve seen as well. So, you know, concerning but the ability to sort of predict that at the point when you are back, then, you know, that’s great news, we just have to sort of enable more of that. My sense is, we’re going to see more of this from you know, companies like yourself capabilities like yourself, because we’ve learned so much more posing minute, where do you see this going? What’s next?
James Lu
Just on the COVID side, specifically, I do think there’s an incredible opportunity being generated right now around surveillance and the new vaccine technology. And the idea that you could surveil a new variant, understand its phenotype. So that’s clinical course, understand that escapes vaccine and then get a new vaccine into market, and kind of the month timeframe. I think it’s not only transformative for, you know, COVID, but for all future sets of infectious diseases, and so it’s been, you know, obviously, it’s been a tragedy the last 18 months, but I think the the deployment of us, and certainly worldwide expertise and Life Sciences has been incredible. And watching it in action and moving so quickly, I think has been tremendously rewarding and certainly for me, like incredibly optimistic about where we’re headed. So I think deployment of new sequencing technologies coupled with these new New Life Sciences technology, I think is everywhere being much better place here in the next couple of years.
Nick van Terheyden
couldn’t agree with you more, I think, you know, it’s terrible to talk about silver linings was such a tragic sort of experience. But that’s exactly what I get out of this. And I agree 100% unfortunately, as usual, we’ve run out of time. So it just remains for me to thank you for joining on the show. Thanks very much.
James Lu
Thank you, Nick. And I really appreciate the time and the interest
