Salisbury: All right, so we are just going to dive in. My name is Meredith Salisbury. I work with a life science communication company called Bioscribe, and I’m also with Techonomy Bio. It is a pleasure to be here with my fabulous panel, and we’re going to have, what I’m hoping it will be, a very lively debate about ownership and access and other issues about your genetic data. And this is not scientific, this is going to be what you need to know as consumers. So I will just go down.
All of these people have great bios that are on the Techonomy website, I encourage you to look at them. I’m going to focus on why they’re here today. Linda Avey, in a past life was a co-founder of 23andME, is currently working with We Are Curious. She has spent several years really focusing on getting consumers direct access to their data. Elissa Levin is at the Icahn School of Medicine at Mount Sinai and is a genetic counselor. One of the rock stars in a very small field. She’s going to talk to us about how people who are actually trained to interpret this data look at the world. Hank Greely from Stanford is here to be our word of caution and reason, which he may not have known until right now.
Greely: That’s quite a challenge.
Salisbury: Matt Wilsey is with the Grace Wilsey foundation and he’s here because, as a parent, he tracked down his daughter’s ultra-rare genetic disease. And when I say ultra-rare, I mean he has spent a few years now going around the world identifying people with this disease, and you can count them with your fingers and toes. And Ajay Royyuru from IBM has worked with the National Geographic project, and really focuses on privacy and security of this data and what you can do.
So we’re going to dive into access. There are a lot of debates right now in biomedicine about consumer access to genetic data. And the spectrum is, you know, people should be able to get their genome sequenced or a test or anything and just get whatever access they want, to the other end which is very regulated and you shouldn’t be able to see any of this data without a medical professional guiding you and interpreting it for you.
And actually, right before we dive into that, just by show of hands, how many people here have had their genome sequenced? Good, very cool. How many people would like to get their genome sequenced? Okay, excellent, you’re in the right room. And then, how many of you have any kind of genetic data, either from a clinical genetic test or a consumer genetics test? Okay, very good. That gives a sense of who the audience is.
So Linda, I’d like to start with you on the access issues, since you are so strongly on the direct access side. Tell me a little bit about why people should be able to get their data on their own, and why there should not be an intermediary.
Avey: Well, you know, part if it is just—I think it’s just logical that people want to learn about themselves and they should have the right to be able to do that. And I think done in a careful way and a responsible way, where you’re making access to the right people who can help navigate questions that you might have. But that first phase of deciding I really want to do this and I want to be able to do this without having to go through my physician. I think that just should be a right that we all have. I don’t think they should be the guardians of that information.
I think that there are other channels, and you know certainly that was the idea for 23andMe, was that you know, if you can provide this interface to help people start to learn about what this all means—we’re still really kind of in the early days of truly understanding what is going on in the genetics research world, and to be able to provide an interface into that that goes beyond what you might read in the “New York Times,” you know, this is something that gives you a little bit deeper understanding. But, you know, I think there’s often this sort of knee jerk reaction that people can’t understand this, and we really wanted to challenge that, to say, you know, if you put it in the proper context and you really do your work on building an interface that people can navigate, we felt like that was very doable. And you know, having conversations with physicians, when you kind of positioned it to them that, okay, if this is going to have to come through you, then that means you’re going to be responsible for the delivery of that information and you’re suddenly going to be put into that position of having to be the explainer of it all. So I think in looking at it that way, and certainly with genetic counselors—you know there aren’t, like you said, it’s a small community of people.
The other thing that we learned at 23andMe is that a lot of people would get their data back and they’d be like, “This is so boring, I didn’t learn anything that interesting.” Right? So for them, they kind of read through the reports and got through it, and it was like, “Fine, move on,” and there wasn’t really that need for—they didn’t have a lot of questions that came out of it. So to have a person situated at the right time if those questions come up, that was really what we were striving for.
Salisbury: And there is a big division between genetic information that’s going to be medically useful and just genetic information, right? And 23andMe, you know, really worked hard to differentiate those, right? Those are the kind of recreational, “Do you taste bitter” genes, right? Versus, “Are you going to get Alzheimer’s?”
Avey: And that again, kind of going back to that interface of building that in so that, you know, hopefully people can read something and get a pretty good feel for how strong these genetic components are, what that means from a risk perspective. We really thought long and hard about how to show whether something is really solid genetically or if it’s just a new research discovery. So that was where the 4, 3, 2, and 1 star reports came from, just to try to give people—because a lot of times the media would report on a discovery, but people typically in journals, even science journalists wouldn’t necessarily go into the actual paper. You know, read how big the patient population was and what the statistics are behind that particular finding. And so that was something we felt like, if we can put it out there, regardless of whether it’s a new discovery, an old discovery, or one that’s really solid or something that’s early on, as long as you explain it in that light, then hopefully people can understand that.
Salisbury: I’m going to jump around and I’ve told all of the panelist to just ignore me and question each other and have their own—
Salisbury: So I fully expect to lose control of this discussion in about two minutes. Matt, you and I were talking the other day about trends in access. Not what it should be, but what it is when you experience it through the medical world. So tell us a little bit about getting the genetic information for your daughter and what you’ve seen.
Wilsey: Yes, and I think, you know our story, we were so early before whole exome or whole genome was even really available to a patient or to a family that’s trying to find something for a loved one. And initially we had said, “Hey look, we want to do whole genome sequencing.” We had struck out for almost three years trying to identify what was going on in my daughter Grace. And the clinicians basically came back and said, “There’s going to be very little that we’ll be able to decipher from this. There’s just so much information.” And being a former tech entrepreneur, I knew that I wanted all the data. I wasn’t sure what I was going to do with it, but I knew over time it was going to be useful and if we weren’t able to identify what was going on at the start, we would be able to identify it in the coming years as the search capabilities became better and more efficient. Basically, the more you put into it, the better the output it.
And were very, very fortunate with the people that we selected to do our sequencing. We worked with Mike Snyder at Stanford, with Richard Gibbs and Huda Zoghbi at Baylor, and with Eric Lander at Broad. So these were sort of the cream of the crop. And because they had really never done this before, it was almost like we had gotten in before the door had closed. We actually would be in the meetings with each of those PI’s and the bioinformatics experts like Atul Butte, or the clinical experts like Greg Enns and Bill Craigen at Baylor, and we would literally go through the variants one by one. Of course they had sifted it out to a reasonable number that we could tackle. But, does Grace have this, does she have that? Is there a family history of diabetes? Are there different things that they could pick out, different clues? And based off of that, we whittled the list down.
What was really fascinating was to have a seat at the table, and it wasn’t just an observer seat, we were actually active participants in the discussion. And because of that, we able to identify a novel disease that had never been found in humans before. There was one other case reported by Duke, but for the most part this was a completely new disease that would never have happened without that open collaboration between the research geneticists, the clinicians, the bioinformatics experts, and that patient or family.
And since that time, I’ve really kind of noticed a disturbing trend, which is a lot of these centers are not telling patients and families what they are finding, and that has a material impact on their lives, but then also, family planning. So families thinking that they might have a de novo mutation, they’ve been searching for sometimes 10 plus years. And they don’t know what caused, maybe, their first child’s problem. They had two other healthy kids, they think its de novo. Some of these medical centers give you sort of the inkling that it is de novo. And they go out and have another child that actually has the rare disease yet again.
Just having been in the rare disease community now for about five and a half years, this is happening more and more often, where these centers aren’t telling information. And there’s not really an action from that outside of, “Warning sign, this probably isn’t de novo. Think about that when you’re doing family planning, or think about that when you’re considering a treatment.” And so, it’s almost like the legal, the lawyers have gotten involved in some of these centers that have really done a disservice ultimately to the do no harm mantra, which I think works in two different directions.
Salisbury: So Hank, tell us a little bit about that. Have the lawyers gotten involved? Is that what we’re seeing?
Greely: A little bit. We tend to get involved whenever there’s money.
Greely: But even when there is and when there’s money to protect, as there often is for academic and other medical centers I think your story is disturbing to me, because at the research, diagnostic odyssey, mystery kid level, I don’t think the lawyers should be involved in limiting what kind of information comes through. I do think that in the more consumer context, not necessarily the lawyers, but somebody should be involved in limiting what information comes through. I’m an academic. My bias is information is good, more information is better, but in the real world that’s not always the case. People sometimes make bad decisions based on their misunderstandings of the information they get.
I was quite happy to see the FDA take the action it did with respect to 23andMe. You can still get your genealogical information, you can still get you earwax information. I love the little thing about the blue white, gold white, black dress that they just did, where they found there was no strong genomic association with what color that stupid dress looked like to you.
Greely: That’s all great. Health is different. You look around our society. There is not an industry that is as heavily regulated as the health industry. You practice medicine without a license, you go to jail. You sell a new compound as a drug without approval, you go to jail. You sell a medical device, at least a risky medical device, without approval you go to jail. It’s not true in say the software industry. This is the most regulated industry and I think there are reasons for it. One is, it’s really complicated. Well, software is complicated, too. The other is the stakes are really high, which they are in some software contexts but not necessarily in other software and hardware contexts. But I think the thing that’s biggest is we all get sick. There’s 315 million of us. We are not all college or post-college educated. We don’t all live in the Silicon Valley bubble. We’re not all good at figuring out what percentages mean. Witness, lotteries still flourish. I mean, if people were better at percentages there shouldn’t be lotteries.
I have a family reunion every year of about 120 people, spread, you know it’s a little shifted, but up and down the socioeconomic status, and most of my relatives barely know how to spell DNA. All they know about it is it’s magic. And telling people details about things that are magic can cause problems. I don’t think the problem is likely to be that somebody goes out and gets a prophylactic double mastectomy without further work. Surgeons and their lawyers should help limit that. But I do worry that a women who tests negative for a BRCA1 or BRCA2 test that involves only a handful of alleles, that has lowered her lifetime risk—the test result has lowered her lifetime risk of getting breast cancer from the 12% American female average to about 11.98%. If she says, “Hey, I’m okay, I don’t need to get a mammogram anymore,” that could be a fatal decision. So I’m in favor of regulation of medical information, in a complicated way.
I was very happy also with what the FDA did recently with 23andMe. They approved its application to test for a rare autosomal recessive disorder called Bloom syndrome. They’re not actually testing for the disease, it’s pretty obvious when a kid has it, but they’re testing for carriers, so two people will know that they have a chance of having a child with this syndrome. They said first, the science on this is really strong, we understand what’s causing it, we understand Bloom syndrome pretty well. This isn’t just snip chip association stuff. This is solid science. Second, we think in the carrier testing context, just like over the counter drugs, just like aspirin, or just like a home pregnancy test, the company has convinced us that customers will understand this well enough to make good use of it. They’ll know when to use it, they’ll know what it means. They’ll know when to stop using it, stop relying on it. The OTC, over the counter standard. I think that makes sense. Otherwise we risk across the whole population, not necessarily across the several 100,000 people who’ve signed up for 23andMe, the enthusiasts and the early adopters. But across the whole population we risk doing some significant harm.
Salisbury: And Elissa, I want to bring you in. You were trained in a profession that, by definition, is supposed to be inserted between people and their results. Yet, you also work for an organization where getting data to people is really important. So, tell me how you feel about this, and also about the limitations you see for genetic counselors who make the decisions not to reveal all of the results of what they get back.
Levin: So, yes, as a genetic counselor, we’re really highly trained in not just the science and the medical genetic side, but also the bioethics. I spent years working in academic medical centers and traditional practices and kept coming across frustrations with the way people were getting information about what’s going in their family or not getting information about what’s going on in their family. And it was really frustrating. And hearing over and over again how patients just wanted access to their information. They wanted to be proactive, you know, “I want to know about my health risks, I want to know how to be proactive and preventative”—is really sort of why I shifted from working in that more traditional setting to working in the direct to consumer world and personal genetic testing world.
What I have really tried to bring into that sphere is an element of how can we do this responsibly? What I’ve done in a couple of different scenarios is trying to create a genetic counseling support network for people going through direct to consumer, direct access genetic testing. This has just been a fascinating process, because, what you know from working in rare disease genetics and other scenarios is that a lot of these individuals are really resilient. You know, people can handle information, and seek it out. And now, you know, there’s a lot of different—Hank’s point, I think—socioeconomic access questions that I think we need to address.
But being able to pair the delivery of genetic information with somebody who is professionally qualified to help you navigate what that means, I think is absolutely critical to the responsible and successful integration of this. And part of that is because most physicians are not familiar with genetic information. It’s this huge gap, they don’t understand the value, it’s a liability issue to them, people walk in with their results from 23andMe, or Navigenics, or wherever they’ve gotten, and most physicians, and I know this from experience, just say, “Well, that’s meaningless, we don’t know what to do with that, just don’t worry about it.” So that doesn’t do anybody a service, because we really want to empower patients and there is a lot of valuable data.
But I think that, taking a little bit of a step back, as a genetic counselor, the biggest, most critical issue to me is pre-test education, expectation setting, and consenting. And I think that we can do a much better job. You know, we’re rooted here in Silicon Valley right now, and we need to marry technology, and all of the potentials of technology, around education and personalizing the pre-test education experience, right? What are you going to get out of this? Is this research grade information? Is this clinical grade information? Is this something you can actually use to make decisions in your healthcare? Is this something that’s really just more fun that you can play with and stick into databases that other researchers are using? This is not something that I think is obvious to most consumers, and there’s a real danger if you don’t make that clear distinction.
And I think there’s equally a huge value if you understand why somebody is coming in to doing genetic testing. What are the motivations? Everybody comes in with their own reason. So having worked with a personal genetic testing company where we were offering dozens and dozens of different results and health risks, you get on the phone and speak with somebody and say, “So why is that you did this testing, what were you hoping to find?” And sometimes it was because of something they heard in the media, they read the latest article. They have a family history of diabetes. They had a personal history or risk of a scare with cancer. You know everybody comes at this with a different personal story, and really wants to focus on different elements of the results or the information that they are getting out of it. And I think that there’s a real need to be able to personalize and contextualize any kind of data that we do give consumers access to, because that’ll really help shape how we can do that responsibly, how we can start to help integrate that and prioritize what they’re going to share with their physicians, what they’re going to share with they’re family members. And again, really creating some of the technology platforms around it to try to facilitate that process.
Greely: I want to really strongly second some of that. I think we are going to have a real hard time figuring out how to explain this stuff, both before the test and after the test to normal people. Or to even maybe below normal people. We’re going to have a hard time teaching them how to do it. And they’re just aren’t enough genetic counselors. They’re 3,000, under 3,000 in North America. There aren’t enough doctors who understand this stuff well. I hope we can always keep some personal contact. But I think we need web based, video based—I think we need the video game folks in it, we need Hollywood in it. We need a lot of effort at how to explain this stuff to real people, both before they get tested, and after they get tested, in an effort to minimize the kinds of misunderstanding that can lead to tragedy.
Levin: And I have to say, just to follow on to that part of what has been really exciting to me coming from working in industry over ten years in this space, to going and working in an academic medical center again, is that this is exactly what we’re trying to focus on, and we are trying to bring in, in a more responsible, but innovative way, you know, how you can digitize or create some more electronic genetic counseling features. It’s really an amazing place to be when you see the marriage of some of the industry and the academic sectors coming together.
Greely: And by the way, I think 23andMe did really the best job I’ve seen of just a web interface explanation. It’s not often I say nice things about 23andMe.
Avey: I’m going to go leave now, I’m happy. No, but I was just going to say that’s, you know, the inserting of videos, and when you can do it in a very personalized way—the thing is that you don’t have to create a video for every person, because a lot of that information is going to be the same. You know if you find out your ApoE 34, you know, you’ve got a higher risk for Alzheimer’s, put a link to a video, you know, of a genetic counselor talking about that. And we’ve actually found through some of the work that we had done, is that people actually preferred a video, because they could back it up. They’d be embarrassed to say, “Could you explain that five more times?” But with a video, they can just replay it, and do it in the comfort of their own home. So there are all the benefits of having a web based interface so that people have the opportunity to spend more time and to get to understand it. And you can do this is a very personalized way that, “Oh, because you have that genotype, you get the link to this particular video clip.”
Greely: I’d just say, have them watch part one of the video before you give them access to the information.
Avey: Right, exactly.
Greely: And say, they don’t get access until they’ve seen the video that sets them up for it, and then they can get the video to follow up on their particular genotype.
Avey: Exactly, yes.
Salisbury: I want to shift gears a bit to privacy of this information, because, we’re all very excited about getting it, and then, once you have it and you can’t protect it, what happens then? And Ajay, in your session this morning, you made a great point about the identifiability of this data. Do you want to comment on that?
Royyuru: Yes. So, first I think this notion of, if the sample came from me, I ought to have the information. I may be pushing it a little bit with respect to the other panelists here, but, it came from me. If I want to know it, don’t prevent me from knowing it.
Royyuru: If I choose not to know it, then don’t put it in my face. So you’ve got to respect that part also, right? So this goes to the title of the session, who owns it? The sample originated from me. I have a claim to knowing what came out of that sample, and I will decide what I want to know out of it. Do I want it to be identifiable? Do I want it to come back to me? Or you just take it and mix it up and aggregate it with everybody else’s. Do whatever you choose, don’t bother me. But that sort of decision should be left to me, right? As a—that’s all part of the informed consent process. If you inform me that these are the choices available to me, and what are the consequences of making any of these choices, then to the best extent possible, I will make that informed decision.
And I should then be given also the ability to change it. Some changes wouldn’t be possible, so you know, anonymizing after the identity had been revealed would be impossible. But, going the other way, which is I initially consented to not be the identified. But then I encounter a situation, either with myself or my family where I say, “You know what? That stuff that I gave my sample to, I want that information coming back to me.” So if I’m given an option of being able to re-identify myself into that study, or you know, as long as there is a pointer that is kept, they can use the pointer to come back to me, I should have the ability to change my consent, right? Then I think it’s very participatory, it’s not forced on anyone, so that’s sort of on the consent side.
But privacy, I feel that we are being very binary, in whether information is private or not private. Whether we consent for some use, or no consent at all. I’ve actually not seen—I’ve consented for a few studies, and I’ve not seen a consent form that says, these are the grades of consent that you can provide. It’s usually presented as checkbox, yes or no. You consent, and you’ve already taken the trouble of going to the clinician, or have been introduced to that study, or whatever, or you’ve bought the kit already, and now what? If the choice is between yes and no. I’m going with yes. I’ve no choice really? So I want to flip this into, give me the more nuanced way in which I can be informed and I consent to it, and that includes consent on how much of this information gets disclosed and to who else. That’s the privacy part, right? So, you know, if there is an extremely rare genetic variation that’s unique enough that there are a handful of people on the planet, I don’t need to know more than that one snip. I know that one snip, and maybe gender or something, and I’ve identified the person. So this notion of identifiability should not be binary in my view.
Salisbury: And for people who don’t know what a snip is, so we’re talking about a single genetic variant that could identify you from anyone else in the population potentially.
Royyuru: Right. Or on the other hand, I could have a genetic variation that marks me in an extremely large group of 10 million people. Then the fact that I have that doesn’t mark me—it reduces the population size from 7 billion to 10 million, but that hasn’t really compromised my privacy to a remarkable degree. So every such genetic data element, my premise is that it must be accompanied with some such measure of identifiability. I’m not trying to actually address how we would do it. That is, I think still an open technical challenge. I haven’t seen elegant solutions to that yet. But I believe as technologists this is something that is addressable. You know, because—
Salisbury: Opportunity for attendees, right? Opportunities for this study?
Royyuru: So if identifiability is something that we can measure, then I can consent for others to use my data only to the extent of my comfort of identifiability. Okay? I’m giving you consent, but you don’t have the consent to actually come back and put my name against it. As long as I’m one in ten million, do whatever you want. But if I’m one in five, come ask me before you can do anything with that data, right? And that’s the degree of consent I want to provide. If I have a measure of identifiability, there’ll be all these applications that will come of it.
Wilsey: It’s almost like a control panel where you can turn on and off levers for certain access, and let the individual make that decision.
Wilsey: You, know I want to share this with NIH, but I don’t want to share this with, you know, pick your medical center of choice.
Royyuru: I think another point, actually, about privacy, which is, what I consider to be consented by me may not be consented by my sibling. You know, my sibling or my very close family, automatically their information is getting revealed. You know, so I go to 23andMe or I go to Genographic and I type myself, and I say, “Hey, wait a second,” you know, “I’m so and so.” And guess what? I’ve identified all my close family members whether they like it or not. And that has actually compromised their privacy, you see?
So this actually is a responsibility that the participants have to take, and I’m not seeing an informed consent process actually educating people well on this issue. So people go blog about, hey, you know, I got tested, I have this and that.” But guess what? You know that actually has revealed so much about your close family members.
Salisbury: And so Ajay, you keep calling it my data. And one of the reasons we’re in this room is to talk about ownership of the data. Is it my data, or do I have kind of a weaker claim than I think. And so I want Hank to help us really think through this process. So, we’re not talking about the genetic material, you know, my arm is my arm and you can’t have it. But the genetic data, right? Because in almost every case, some other company has to generate that data for us, right? So who has the claim on the data? Do I have the claim because it came from my cells, or do they have the claim because they actually generated the data and generated the interpretation of it? So, Hank, where do you stand here?
Greely: Very straightforward answer: yes, no, and maybe, all at the same time. There is no clear answer to this, and even with the materials—the reason I’m carrying this is tomorrow morning at 7:30, I’m getting a new left hip. I’m going to be very happy to see this left hip go. This left hip has served me well until recently, but it’s passed its sell-by date. This is my left hip. You can’t take it out of me without my permission. I asked the surgeon, could I—you know, he’s going to cut off the top of it—could I get that, the femoral head of my hip and put it in a jar? And I could hear him over email laughing at me. “No, of course not.” Even things that clearly ours in one context become medical waste and we’re not allowed to have them in another context.
Ownership—I teach first-year property students among other things, first-year law students, property—and one of the things we do is try to point out how tricky the concept of property and ownership is. Ownership isn’t a nice single state—I own my car, but I’m not allowed to drive it more than 65 mph, in theory. 80-so on 280, on a good day. But ownership has lots of limitations. The dominant metaphor for ownership in American law in the last 100 years has been the bundle of sticks—that it’s a bunch of separate rights and duties, and they get over a certain size, and we say you own it, but that doesn’t mean there aren’t other people who’ve got some aspects of that.
With respect to data the answer is really unclear and not very satisfying. It’s not clear what it even means to say it’s owned—you have the opportunity to authorize somebody to do the things that extract that data or not. You’ve got the choice to consent, most of the time—although if you’re a newborn baby, you don’t have that choice and your parents don’t have that choice, with respect to 30 genetic diseases—for a bunch of good reasons, actually. But once it’s extracted, who owns it? The hospital, the medical center will say, “Well, it’s in my chart, I own it.” If it’s through an NIH study, the NIH will say, “Well, you agreed—when you got your NIH grant, you agreed to share all the data and put it in dbGaP, the Database for Genotypes and Phenotypes,” which NIH will then let any researcher who makes a plausible story get access to.
Who owns it? Oh—anybody here have a clone? Anybody here an identical twin? The odds for the group this size are pretty good that somebody would be. Nobody? So if you had an identical twin, who owns your genome? Because she or he owns your genome too, same genome. As well as the fact that all of your sibs, parents, and children own half your genome, so it’s not just ownership is tricky, but “my” is tricky.
And the answer right now is once somebody has gotten it and analyzed it, ownership of that data is very much up for grabs. I’d say the best arguments—and they’re not very good arguments, are contract-based arguments, based on the consent forms, except that most universities and researchers and doctors take the position that consent forms aren’t actually contracts. You have do what you promised, but they don’t have to do what they promised.
So the answer is really obscure. Yes, no, and maybe.
Greely: I specialize in taking the confusion to a whole deeper level, yeah.
Salisbury: And that is why you’re here. Matt, I wanted to ask you about the data you had generated for Grace, and for your family. You got it from various sources.
Salisbury: What did you do to, you know—gather, not gather, protect it?
Wilsey: Yeah, we got all of it. We have hard drives with it, we have secure FTP sites that we’ve shared it with other research centers. We’ve been talking with Google, for example, I know David spoke earlier today about putting it in the cloud. We are big believers—I appreciate everything Hank said, but we are big believers that it is our data, that we own it. We sought out researchers specifically that had that same sort of vision, and shared that same philosophy. Now, at least, that was just in our micro, our little microcosm.
It’s probably evolved ten times since then, since we’ve done it, and their views have probably evolved as well, but at least at that snapshot in time—and maybe we were just very fortunate, right place at right time—but they have been incredibly open and accommodating with the data, and if a new researcher comes in, I will call Mike Snyder and say, “Mike, can you please prepare a hard drive and send it to this person for me?” Absolutely, and they do it.
And I guess this also goes to the nature of being an ultra-rare genetic disease, where there’s only 30 identified cases in the world, this is really, really rare. Rare disease is anything that affects 200,000 Americans or less, and that’s a lot of people. So imagine when it’s really only, like, 15 Americans. And so maybe we’re cutting some corners that would otherwise be much bigger issues in Parkinson’s or cancer.
But thankfully, and I can’t say that enough, without their openness and collaboration on sharing this information, we would be still in the Stone Ages with this disease, and I actually think that we do have a real shot of curing it, which is just sort of crazy to even say that or even think it, but I think it really is a real shot to do it.
Greely: So, I think in your situation, I would feel exactly the way you do. And I, frankly, think that we overstress a little bit about genetic privacy and DNA privacy. I’d much rather let you have my genome than let you have my credit card records or my Google search records, which somebody has—probably several people and entities have. But it’s good that they’ve been willing to share when you’ve requested, and that’s part of your deal with them—contractual or quasi-contractual—but you might worry about who else they’ve shared it with that you haven’t told them to. And some of that would be, yeah—Mike Snyder is my courtesy chair, I have a courtesy appointment in genetics, so he’s my chair by courtesy—I don’t think Mike would do it, but you know, if there’s a post-doc or a grad student in his lab who’s worked with that data and then gets a job someplace else, that data is probably going off with that person. And then from that lab, where it goes and where it goes—controlling, as we all know—controlling data is really hard.
Wilsey: Slippery slope. And Hank’s absolutely right, because we do worry about that, where, you know, you want to at least know who has their hands on it, and you can even extrapolate that further, where people are now deriving IPSCs, so stem cells, that we heard about earlier—we’re making different tissue, we’re sending fibroblast, which is skin tissue, we’re sending liver places. I mean, it is like wildfire, it’s all over the place, and quite frankly, it’s hard to manage all that or even know who has their hands on it at any given time. We’ve made the leap of faith that—not just for Grace, but for other rare diseases, and then maybe in our own lives, and how we might develop other things, like Parkinson’s or Alzheimer’s or more common diseases—we’ve just kind of taken the plunge and said, look, at this point we’re essentially an open book. We can’t pretend that we’re not who we are and that that information is out there. We hope that it’s powerful for the greater good, and I think there’s a lot of people who feel that way, especially in the rare disease community, but others who have been sequenced that they feel that this information is very, very powerful, and it should be, in a way, open source. At least the people in my network, which I know is a lot of selection bias, but it’s just kind of an interesting thought, of that, this isn’t just about me, the patient, or my daughter—it’s about how can this be useful for society?
Avey: Yeah, and I think that, you know, the PGP Project, that George Church did, I think he very intentionally just put it way out ahead, by having everyone consent that their data would be open to anyone that wanted to look at it. So that he was very clear about that, that that was the criteria for joining, that you had to agree that your genome would be out there. And he tells a very funny story about how—and everybody’s probably heard it—but of how he had put his own data out, and a doctor, someone happened to be looking through it—and his medical records also were open to be looked at—I think he might have blocked a couple things—but what they noticed, in looking through his data, that he probably could be on a better, I think it was a statin or something. But just by looking through his medical information, they actually sort of diagnosed the fact that he should switch his medication.
And so he, that’s sort of the example on the flipside of being open and the potential value that comes from that. So I think we always have to weigh both sides. We can’t always be the fear and loathing side. I think we have to strike that balance to see—and again, going back to giving people the option, you know, you might want to put some of your snips out. You don’t want to put them all out, but you might want to expose some of them, because you might get some benefit back from that. Or if you have a very rare mutation, you know, maybe we can create marketplaces for people to say hey, I’ve got this very rare marker in my genome, and is that valuable at some level, you know? So that—I think there are just so many things to explore here, beyond just ownership, but utility and value.
Levin: One other thing that I want to add is just, I think we’re at the early stage of an inflection point around really the culture of participating in research and genomic research is obviously one huge element of that. I know there was discussion around ResearchKit earlier today, and being able to engage in an app-level in different diseases, and we’re certainly very invested and involved with that, but I—there’s also a recent, I think just this week OpenHumans.org launched, which is sort of taking the Personal Genome Project in some ways to yet another level, where you can actually—it’s making the assumption you own any of your biological data, and you can actually upload it to this site, you can select what research studies you want to participate in, and you control some of the elements, in terms of privacy and with whom you share it and when you share it. And I think this is just, again, a snapshot of a direction we’re going—and again, this is focused on research information and not necessarily clinical information—but I think it’s a way to start to socialize and get people more comfortable with having control over your information, but again, making that clear assumption that you own it.
Greely: The inflection point can go the other way, which is what I worry about. 5.4 million blood spot samples were incinerated in Texas because parents discovered that some of their kids’ blood spots were being used for research without their knowledge or permission. They sued, they got state law passed, 5.4 million irreplaceable samples, gone.
Salisbury: And we could keep going for an hour here, but I know that you guys have some questions, so I want to leave a little time for that. Hi Jeanne, this is my crew behind you. Please introduce yourself.
Loring: Jeanne Loring, from the Scripps Research Institute. Hi, Hank.
Greely: Hi, Jeanne.
Loring: Hi—some people I haven’t seen in a little while. So just to let you know what kind of person I am, when Allen Roses published his work on ApoE in 1993, within two weeks I had amplified that gene and did the restriction digest and discovered what my ApoE genotype was, which is 3/3—I’ll share it with you all.
Loring: But I want to point out two things that 23andMe had huge impact on us—we were early adopters. And so we still have access to all our health data, and we have 12 members of our family who were also analyzed early enough so we have access to their health data. So one thing that happened is, my husband has an adopted daughter, and she was planning to have a baby. We had her snip-genotyped, we had her husband snip-genotyped, we discovered she had a cystic fibrosis recessive mutation, and so actually we did her first, and then we did her husband to make sure he wasn’t also. She’s adopted, there was no family history here at all, so we had no way of knowing that she might have inherited this, so for adopted kids, I think having a snip genotype is the same thing as having a family history.
The other one is more personal—it’s the same as George’s. So, I found out through 23andMe that I have two of the variants that are associated with the really awful side effects of taking statins—so, which is very common in Europeans—so when my doctor said I should take statins, because my cholesterol had gotten up high enough, I said, “Well, you know, I’m not really sure about this. This doesn’t seem like a good idea.” But she talked me into it, and I said, okay, I’ll do an experiment. So I took Crestor, lowest dose, twice a week for, off and on, for about three months. I had the worst muscle pains. This is exactly, this is exactly what the symptom is. My doctor was really impressed, and now she’s gone back, and she’s going to get her other patients genotyped, because she has patients with such a severe side effect that their muscles are actually atrophying. And wouldn’t it have been nice to know they shouldn’t take statins, early on, right?
So I think more information is better. I disagree with you, Hank, in that I’ve made IPSOs of myself, I’ve distributed them to people, I’ve given out my sequence. So the more, the better, as far as I’m concerned, and I’m willing to be an experiment, you know? If something bad happens to me, I’ll write about it.
Greely: I think we can safely say the entire world is not made up of George Churches and Jeanne Lorings, though. A shame. Kind of.
Meredith, there’s one here. Speak loud, we can repeat your question.
Salisbury: Hold on just one second, we’ll get the microphone to you, yep.
Perec: Hi, Nimoy Perec, Bell Labs Technologies [PH— 0:45:37.6]. I was curious to know that a lot of mentioned that there’s certain snippets of information that are more useful than others, and that the public should know about that. And I was wondering, what would be the best strategies to use, so that people actually know what the ramifications are for knowing which snippets to put out there and which ones to obviously keep closer to the chest?
Levin: I think that there’s a lot of debate amongst people who are in my profession, the medical genetics world. We use the term actionability a lot, right, so which genetic variants are medically actionable or, you know, clinically relevant, and there’s a huge debate right now. And so, while we would all love that to be a really simple answer, even taking a relatively conservative approach, and saying, okay, if we sequence your genome, we’re only going to give out risk information on genetic mutations that we find in these 56 genes that are related to these particular diseases.
Now, every institution that’s offering this kind of testing or service, might expand that—obviously different consumer-centric organizations might want to extend that even further, around pharmacogenetic testing, and traits, and other ancestry-oriented testing, but I think that there’s just a huge spectrum in terms of what we understand. Yes, this variant might be related to this particular condition, but what you can do with that information is, I think, a wide-open topic for discussion. And so I think part of the answer to that is really rooted in what are you trying to get out of that information, right? What are you expecting to find, and having some sort of conversation about that, because there is no clear-cut black and white, you know, this is the useful part of your genome and this isn’t.
Greely: Although there are some parts that we know—and it’s more than the 56—a lot of the rare autosomal recessives, for example—there are a bunch of things that we know pretty clearly, there are probably 1,000 or more genes where certain mutations are known to be strongly related with disease. Most of the genome we know, we are confident, we are real sure—know is a little strong—has nothing to do with anything, or has nothing to do with your particular disease, but the hard stuff is all the stuff in the middle, where there are one or two studies of snip associations that say, “Well, you know, with this snip, people have 20% higher risk of X, the 500 Finnish people we looked at. And in this snip, you know, with the 400 people from American, it’s 10% lower.”
That’s the hard stuff, the stuff in the middle, where you’ve got a little bit of information and not very much. Personally, I prefer more regulatory approach to it, but what we really need to make sure, you know, is to have good data and good databases that tell us how strong the evidence is. Whether the “us” is the FDA making a decision about whether something should be allowed or whether it’s a doctor or a genetic counselor making a decision to rely on something or whether it’s a direct-to-consumer company deciding to put it on its panel. Getting that information is the key part, and they’re working on it, but it’s slow.
Croll: Alistair Croll, from Solve for Interesting and O’Reilly Media. So I spend a lot of time looking at the data side of this and it seems to me like today there’s this sense that no one should know more about you than you do. It’s a good sort of founding principle. But what I’m curious about it, what happens in about four or five years when Moore’s Law catches up and I may not be able to make sense of genetic material or medical conditions the way a doctor can but I can use Watson for an hour for a hundred bucks, upload my data?
Seems to me like, if we look at where technology is going, the ability for machines to interpret this data—it’s going to be very hard for me to get that data, and then say, “Could you please look at this software-as-a-service model for my data,” and that changes sort of the benefits here. We’re saying, look, you can’t be entitled to this information because you don’t have the skills to interpret it, but as machines that we can rent fairly affordably can interpret it fairly efficiently, does that change the balance of whether we should be protecting people from their own data or giving them access to it?
Greely: I’ll take the first shot at it. I think there are two different issues here. One is, we are going to have largely, but not entirely machine—we already have largely but not entirely computer interpretation of genomic data. It gets run through a variety of screens, the things get kicked out, most stuff gets curated, ultimately, by a real human being, but a lot of the initial decisions are made by software. That will get better. It will be a long time before it’s really good, because our underlying data isn’t very good—and garbage-in/garbage-out is one of the few computer things I remember from my Computer Science class 40-some years ago. It’s still true, I think.
So I think that will help with the interpretation. You probably, at least for a long time, want some qualified human in there to confirm it, for everything but the most obvious. But the real thing that I worry about isn’t that individuals won’t understand—the ACGT in this particular location leads, is associated with that disease, but they won’t necessarily have a good understanding of what that means for them, and what they should do. And that’s where genetic counselors, I think, are invaluable, to say, look, yes—you don’t have a mutation in BRCA1 and BRCA2 that will put you at 85% risk of breast cancer, but you still need to get mammograms, but you’re still at 11.89% risk of getting breast cancer, instead of the underlying 12%.
If that, bringing it down to the human level—and with some humans, I think there are people who are very good at understanding and interpreting that, and there are people who aren’t. And whether a machine—I think machines will have a role to play in helping real, average people understand this stuff. I think we’re a ways away from being able to count on them to do as good a job as we’d like.
Levin: I also think one of the huge opportunities in that area is around consumer engagement, right, so the more data we have, the more data we have about you, the more that we’re going to be able to better understand how to interpret that data for you. So I think, you know, one of the areas that we’re really interested in is how do you create, you know, these engaged genomic consumers over time? Because part of that, well, obviously our interpretation at different points in your life cycle, different parts of your DNA are going to have different relevance to you.
So I think, you know, there’s going to be this concept of, the more data that you’re able to share, you know, at a clinical or research level is going to be useful, but then you’re going to want to reinterpret it over time, and in some context, that is going to absolutely involve, you know, a qualified healthcare professional.
Salisbury: Ajay, is Watson going to solve the terrible shortage of genomic counselors?
Royyuru: So, you know, we are looking at that. Not with germline mutations at this point, but really somatic mutations in cancers, which, you know, cancer is a progressive disease, so later stage of cancer, you actually have an extremely large variation files that describe lots and lots, thousands and thousands of mutations. So it is an interpretation challenge to translate all the raw genomic changes in the cancer biopsy to what is the actionable insight, you know? To mechanisms of action, to if there are drugs that are possible for this patient, which of them have what sort of outcome in patients who have the following genomic mutations as well. So that sort of is an interpretation challenge and we think of Watson as a cognitive assistant that can assist in doing this task, bring the attributes of comprehensive knowledge transparency to the decision-making or the decision support, and actually do that in an extremely short period of time.
This actually is a task that genetic counselors and clinical oncologists are highly skilled at doing but we don’t have enough of them to treat all the cancer patients with the speed and rigor that we need.
Avey: But I would challenge that even if you had enough genetic counselors, to do that kind of analytics is—why would you have a person doing that, you know? I mean, that’s just—we have computers, that’s what we should be using them for, and it’s a combination of the computer spitting out—after crunching all the numbers, looking, comparing your tumor against all the tumor database—that’s just something that, humans aren’t suited to do that. But then having that information come out, and then have a doctor or an oncologist or a medical geneticist look at that and say, okay, this looks like the appropriate treatment, but let me talk to some of my colleagues and let’s—you know, so it’s that combination of computer with human, with the intelligence and experience that they bring to it, but yeah, I think there’s so much hope a Watson-like system—
Greely: And then having the doctor or the counselor explain it to the patient.
Avey: Right, exactly.
Greely: And what the options are and what the odds are.
Avey: Exactly, right.
Royyuru: But in a precision oncology context, the way we are implementing it, Watson does the cognitive analysis, present a report, which is primarily consumed by the Molecular Tumor Board—so that’s a panel of experts who can digest that information, contextualize it and implement it for the patient. They’re actually more important to the patient, they’re providing—typically that Molecular Tumor Board are providing that as an advisory help to the clinician, who then talks to the patient. So you’re like two steps removed from the patient, and all of this is actually a way to distill the information, which is very vast, but distill it into more and more actionable insight.
Salisbury: Great, so we’ve got to move on to the next question.
Audience1: I know you’re out of time, so maybe I’ll make the question simple—it seems to me very naïve to say, “Those are my data.” The issue isn’t about the data, it’s about claims. Different than the claim you talked about, Meredith, it’s the claim that your data is associated with a higher risk or a lower risk—that’s when the regulators get involved. And so this recent experience with 23andMe, having a rare genetic disorder that got a claim—my question is about scalability. If you have 500 things or 1,000 things you’d like a claim for, as a consumer genetics company—do any of the panel members know of a concept that makes those claims scalable?
Greely: It’s going to be a problem, and FDA knows it’s a problem. FDA put out a draft guidance on what are called laboratory-developed tests, which include most but not all genetic tests, and expressly didn’t include direct-to-consumer tests. They got lots and lots of feedback. We don’t know whether they’re going to put out another draft guidance, whether they’re going to go forward to something final, whether they’re going to scrap the whole idea. But a big issue for them is how much information will it take before we feel confident that this genomic variation is, in fact, strongly associated with this disease or drug response, or whatever? Nobody, I think, believes it’s going to require clinical trials. That’s just a complete non-starter. How much less than that, what kind of showing you have to make—some sort of peer-reviewed literature showing, I think, is where they’re likely to end up, but how much of the peer-reviewed literature and who interprets it and how—those are going to be really hard questions that they’ll be dealing with for a while. Or they won’t. They’ll just say, “We’re not going to deal with it,” and it goes back to whatever anybody wants to claim.
Salisbury: All right, we’ve got to wrap it up. I just have one more question for the panel, and I’m only going to let them have one-word answers. Right now, in this moment today, would you—your only options are yes or no—would you recommend that the average consumer gets his or her genome sequenced?
Avey: No brainer. Yes.
Royyuru: Not whole genome, no.
Salisbury: All right, well, please join me in thanking our great panel.