 All right, thanks very much, and I'm glad to have the chance to try and fill in some of the gaps, pardon the pun there, from yesterday. And what I want to do is two things this morning. First tell you a little bit about some of the work that we have done in our research program that led to some of these insights that contributed to the debate on incidental findings, and then in the second part tell you a little bit about the incidental findings report itself. So predictive medicine is an interesting business to explore. And it's interesting because it's difficult and it's challenging, and it's something that a lot of people are happy to talk about in the general, but when you start actually talking specifics and making predictions, people get really uncomfortable. And I've learned a lot about our field in doing this, and it has been an interesting journey. So what we learned from is this project we call ClinSeq, which we started back in 2006 before we could even do clinical sequencing practically, and we at that time predicted that we would be coming to this juncture, although to be honest with you we weren't predicting it would come as fast as it did. So we went ahead and started to build a cohort of people that we consented for full genomic interrogation and began to gather them, and we started by focusing on atherosclerosis, and worthwhile phenotype because it's known genetic complexity, and it was basically a positive control for some phenotypes that we wanted to study. And we phenotyped the patients very lightly for that disorder, but then recognized that what we needed to be able to do, and this was critical to the study, is iteratively phenotyped the patients. So that is after we go in and dig into their genomic findings, we take the genomic findings and use that to drive the phenotyping. And that has been something that has led us to some surprising findings and conclusions, and I think it is one of the ways that we'll need to go forward, and also it models how I think medicine can be practiced in a different way. And so we were piloting incidental findings and predictive medicine through this study and began to dig in since we're focused on atherosclerosis initially, then we did not ascertain these patients for other phenotypes, and they can be regarded as an unbiased, ascertained population for other traits. And so what we started with was cancer susceptibility. So we went through the literature, catalogued every known heritable cancer syndrome we could find, and asked the question, how many of our participants have variants in genes known to cause these traits? And the number was pretty surprising, which includes, in fact, more than one percent of our population has a known pathogenic variant in a gene known to cause inherited cancer susceptibility, most of which is heritable breast and ovarian cancer susceptibility, and the variants are listed there. And I will remind you of following on Bruce Korf's comment yesterday, these variants are not in the category of variants that one would wonder whether they cause cancer susceptibility or not. These are very well-established pathogenic variants, and so there isn't really any debate in the field amongst experts who study these genes as to whether these variants are pathogenic or not. And arguably, our threshold is too high, and that we are bypassing variants that are potentially pathogenic because we're trying to set our positive predictive value to be high so that we're not chasing ghosts and finding things that are real when we find them in the patient. So Dan. Clearly pathogenic, are they 100% penetrant? Almost no variant in cancer susceptibility is 100% penetrant, so that is not a realistic metric to target for. The breast and ovarian, the high penetrant ones run in the 80 to 90% range. The Dell 185 AG, I think, is about 60%. No, those are high. Those are high. They're not 100. Yeah. And so, yeah, the relative risk is high for females for those. It's stupendously high for BRCA2 for males. Actually, the relative risk of cancer is higher for men than it is for women with these mutations for breast cancer. There are a lot of other traits we have begun to explore. We have a paper in press now on malignant hypothermia, an important gene environment trait that predisposes people to have a catastrophic reaction to certain anesthetics. Prevalence is stupendously hard to nail down. The estimates are all over the map, and people argue about it incessantly, and one of the reasons is because the ascertainment is so erratic and difficult. But what we can do with genomics is now begin to get a clearer picture of what the actual prevalence is, and we went through our participants and found a known pathogenic variant. Interestingly enough, there's, you talk about standards and guidelines for things like ClinVar. This is sort of a wonderful gold standard for this trait. There are actually two societies, the American Malignant Hyperthermia Society and the European Malignant Hyperthermia Society, and they have developed consensus mutations for variants that they believe are unambiguously pathogenic for this trait. And it turns out that this particular variant that we identified in our participant is present on both groups' lists, and they agree that this is a pathogenic variant. And interestingly enough, we go back and look at this gentleman's family history. He does not have a personal or family history of malignant hypothermia, but he's been exposed three times to anesthetics that could cause such reaction. When you study families who have malignant hypothermia susceptibility, this is what you see. Many of them, they just pop up out of the blue, and it is not with every exposure that the reaction develops. And so this is, we would regard as a great catch of an individual who has the susceptibility, and there are relatively straightforward steps that one can take to minimize his risk going forward, test his family for segregation of this, et cetera. Cardiomyopathies and dysrhythmias are another challenge. These are extremely heterogeneous traits. They are harder. And they're harder because a lot of the variants that cause these traits are missense variants. They're extreme genetic heterogeneity and allelic heterogeneity. And some of the reports are very thin on actual data, and it can be very difficult to make determinations of causality. But the iterative phenotyping really helps here because these are traits that you can get some phenotypic traction on. So we found three variants in the cardiomyopathy category that we were intrigued by, and then went back and either did additional studies or dug out clinical data on these patients, and we found confirmation of the traits in all three of these participants. So the first one has a fascinating disorder called non-compaction cardiomyopathy. The variant predicted that. And it turns out this gentleman had had a previous MRI done for other reasons, and pursuant to the incidental findings discussion, interestingly enough, the radiologist missed the finding of non-compaction cardiomyopathy on his MRI. We went back and pulled it up, and with a focused look at that MRI data, the finding is there. So he absolutely does have this trait that predisposes to congestive heart failure and dysrhythmias and is a manageable disorder. The next two are more straightforward cardiomyopathies, and both of these participants did have evidence of abnormal echofindings consistent with cardiomyopathy. And what's interesting about all three of these families is that they all had a positive family history of sudden unexplained cardiac death. And so they had enrolled in a study on atherosclerosis thinking that they had sudden MIs, or sudden cardiac death, presumably MIs, because that's the attribution that is most commonly made. And we don't think that that's actually what's going on in these families. They've been misdiagnosed, and this trait is what they have. Lester, did you do the targeted phenotyping on the cancer mutations that you found? Because you said you found all those, but you didn't report on whether there was a family history or anything of cancers in those patients. So to be a little tongue-in-cheek about it, you can't really refute our findings, because cancer susceptibility is susceptibility. So for the breast and ovarian cancer susceptibility, we implemented screening for those individuals per the standard recommendations for those traits. But when we look at those patients- Was there a family history? Was there a family history? Half of them, yes. Half of them, no. And that is actually what the oncologists will tell you they generally find when they do genetic testing in the cancer clinic, when patients have disease, when they order testing, half of the ones that are mutation positive, half have a family history and half don't. When we look at our family histories, what do we see in the ones that have a negative family history? Turns out that most of them are either from families with very small, sipship sizes, i.e. there's not a lot of people in the family, or in the breast and ovarian cancer susceptibility families, they're loaded with males. And the penetrance for males being much lower than females has masked the segregation of the trait in those families, but it's there. And several of the participants have female offspring who are at risk, and we've identified them before the disease has manifested. And it's a very interesting phenomenon, too, because we have this paradigm for cancer genetics, which is you go to an oncologist because of either a family or a personal history of this disorder, and you're screened and evaluated to determine if you meet the criteria that would justify the test. Which basically what we're saying to people is, no, no, no, we're not testing you until people have suffered or died from the disease. That's our standard today. And so we're very comfortable with that. There's good reasons for that related to the resources and the positive predictive value and the false positives. But it is actually kind of a funny place to be in. The answer ought to be, well, you know, we would like to find this before people are sick or die instead of after. And so how do you do it? But that's prediction, and it is challenging to do. We're also exploring the sort of the wider space of variants. And there was a wonderful paper by the Thousand Genomes Project where they categorized all null variants in thousand genomes. Really interesting how many there are. And what we set out to do is ask the question, okay, how many null variants are there in our thousand patients, and do they actually have a phenotype? And that's one of the strengths of our approach and one of the Achilles' heels of thousand genomes is you can find a null variant in a patient in thousand genomes. But what does it mean? And there's actually a striking, there is a lack of phenotypic data there of necessity. It was appropriately done, but you just can't go back and ask the question. We can go back and ask the question. Yes, sir. Nonsense, missense, canonical splice, plus one, plus two, minus one, minus two, or deletion of the entire gene. So that's what we consider to be a null variant. And we have to be careful, too. I will tell you also we have null variants in genes that cause nearly 100% penetrant malformation syndromes in patients in their sixth decade who absolutely, positively don't have those traits. Not all null variants are truly nulls, and you have to be very careful about how you interpret those. And we have criteria for how we filter those to make sure that they're in regions of genes that are pathogenic, et cetera. We are looking at heterozygotes. It's very, we have some experience looking for bililic mutations. Mathematically, that's just much harder to do in a population, a cohort of only a thousand people. You're just not going to find that many. We found a patient who had an undiagnosed familial partial lipodystrophy and a PPA or gamma mutation. I was stunned. We found a patient with a hox mutation. Hox mutations cause congenital anomalies, and we thought, sure, this is a false positive, right? And we call up this gentleman and we start saying, well, actually, sir, we have some questions about your hands and feet that cause limb anomalies. He says, what do you mean? You mean the fact that I can't wear flip-flop sandals? I said, what do you mean? Well, he says, there's no space between any of my toes. He has phenotype. We didn't phenotype his feet because we're studying atherosclerosis. If you go back and ask the pointed question, there is the trait. It's not a medically actionable disorder to be sure, but there's an example of a developmental anomaly in a person in their 60s undiagnosed. And we have identified a patient with Burd-Hawk-DeBase syndrome, undiagnosed, negative family history, brought them into the clinic. This causes some particular benign hyperplastic skin lesions plus kidney cancer susceptibility. And we actually had a dermatologist carefully examine the patient, had some suspicious lesions, biopsied, he has the lesions. So finding of this disorder in a person with no personal or family history. We have explored a copy number. I have a question. I'm sorry. What is the specificity of these findings? In other words, do you have subjects that have these null variants that didn't have any phenotype? Absolutely. So we have, I studied years ago, a disorder called Gregg's cephalopolysyndactyline syndrome, which is a disorder that causes polydactyline of all four limbs and craniofacial abnormalities caused by loss of function mutations in Glee 3. And we found a patient who had a nonsense mutation. I think it was in codon 15 of Glee 3, so way five prime. We were stunned. We go back. The patient has absolutely no phenotype. I think what that's telling us is that that is not an obligate exon in that gene, even though we thought it was, because we know what the phenotype is and it sort of unambiguously does not have it. So there are examples of that. And I could enumerate those. I just don't have time to go through them this morning. So you absolutely have to be careful. And that's why we think the iterative phenotyping is essential. You don't just take the variant at face value. The analytic validity of the findings we don't think is our challenge. What's the challenge is coupling it to the phenotype. And that is the core activity we think of this process, which is a clinical researcher or a clinician taking these raw data from the genome and integrating it into the phenotype and asking the question, does this make sense? What's the specificity of the phenotype? Does the patient have or not have the phenotype? And that's what we think is going to have to be done. And that's the core activity that we're going to need to learn how to do. But for a national ability standpoint, I think the specificity of the finding is quite important for sort of point of care or at least preventive measures or other interventions. You couldn't agree more and that for sure when you have a non-refutable phenotype, such as a cancer susceptibility syndrome, then you're in a tough spot and then you're relying more on the actual positive predictive value of the finding itself. We had some variant patients who had a variant in a gene that has been reported to cause gastric cancer. It's really a tough phenotype to deal with. It's very difficult to manage that. We had two patients in our cohort and actually we went so far as to actually contact the research group that published the finding and come to learn that they have since become skeptical of that variant. It was too frequent in databases, honestly, to cause that trait. And they have, in their own minds, downgraded that from pathogenic to benign, but they never published that finding. And so that's an example of how you have to sort of chase things down in the weaknesses in our databases. That's a big challenge. Yes, sir. So you're dealing, in terms of phenotyping, you're dealing, how do you do, one problem you have is age of onset, obviously. Yeah. And also most susceptibility genes are the major contribution to probably most medical problems in terms of numbers, comments. So how do you, you're just looking at things that are specifically causal, is that right? Or how do you handle any susceptibility? So you're talking about lower penetrance and lower relative risk. Yes. So what we're doing is basically taking this data set and starting from what we think is the top end of penetrance and relative risk. And going through those and working our way down the list. And one of the things I think that we will do in the process of doing this is, two of the things we'll do, is figure out sort of where, how that's, what the shape of that curve really looks like by studying these patients and these phenotypes. And the other thing I think we will learn is where the fatigue point is in the participants. Right? We can't keep finding, because each one of them has thousands of variants. We can't keep going back to patients forever and saying, oh, isn't this interesting? We'd like to look at this. They're going to start telling us, you know what, this isn't sufficiently important to bother me with. Would you please leave me alone? That's an important finding. We need to know where that is, because we need to know how far to push this predictive model. And we don't have good data on that. It's a great question. Oh, we have explored CNVs. It's hard to do CNV calling from exome data, as many of you will know. But from whole genome sequence data, one of our participants, and two exomes, we were able to identify a CNV, which is a known pathogenic CNV and a gene called PMP22, peripheral myelin protein 22, which causes patients to have hereditary liability to nerve and pressure palsies. And it turned out when we interviewed these patients that in fact they had these symptoms and they were in all three cases undiagnosed. Interestingly enough, the prevalence of this disorder is supposed to be one in 50,000 people. And we have it in three out of 1,000 people, which is startling. I'm, of course, biasing you because I'm only telling you of our positive findings, not of our negative findings, right? So just get to be careful with that math. But when we explored this with the participants, one of them had actually been misdiagnosed with a lumbar disc radiculopathy and was booked for surgery on his spine because his orthopedic surgeon was convinced that his neuropathy was from his MRI finding, which showed a mildly herniated disc. He canceled that surgery, went to physical therapy and is doing fine. And I think it's a good example of how a correct diagnosis can keep you out of the operating room, save expenses, and save healthcare costs. So what we've shown, I think, is that you can extract clinically meaningful results from exome and genome sequences. Genomic analytic validity to us is not the major challenge. It's integrating the variant findings with the phenotype, which is why we think that the main action is in the clinic room with a clinician who understands both the genomic data and the phenotypic data, understands family history, understands genetics, understands how to integrate all these different data to make a sensible clinical conclusion about a provocative genomic finding. We have become rather enamored with our notion of what we're calling hypothesis-generating research, using the genomic data to tell us what to phenotype and not trying to phenotype patients a priori. And I will tell you that counseling these patients is challenging because in our prior genetic model, what we do, again, is we have patients come to us who have a family history or personal history of a disorder. They have a lot of experience with that disorder. They know what it's about. They know how it's affected them. They've done a lot of the emotional hard work of adapting to that trait, and we take the next step in that process. When you take a patient effectively off the street, sequence them, and say, oh, by the way, you have a disease that you have no experience in our understanding with, it's more startling. There are more challenges, and it takes more time to get those patients adapted to the finding and get them to do what needs to be done to manage that trait. Okay, so I don't want to minimize the challenges here. Predictions are very difficult, and as the gentleman in the background said, they are beyond the domain of exact prediction, not because of lack of order in nature, but because of the variety of factors in operation. Of course, he's talking about particle physics. We're talking about genomics, but the exact same principle applies. We need to understand clearly what the boundaries of the precision of our predictions are so that we can apply them thoughtfully and sensibly to improve the patient's health care. So Rumsfeld was quoted yesterday, I think, by the Admiral. Sort of paraphrase him here. There are things we know are problems and things we know we don't know, and we have to anticipate all of these. It's absolutely clear that clinicians are not ready for this. It's fascinating to return results to these patients' doctors. Some of them do actually kind of wig out when you tell them because this doesn't fit their paradigm. They're used to taking care of patients because they're sick, and we're telling them, no, you need to start taking care of this patient because of a genomic finding, and that is challenging. Databases are an enormous problem. We talked about that yesterday. And what we know from this kind of work is that we don't actually understand the full phenotypic spectrum of most of the traits that we've studied because we've only studied the highly loaded, dominant, recessive, clearly inherited families in the past. And so these kinds of research projects are an opportunity to actually ascertain patients in a less biased way and understand the full spectrum of genotype-phenotype correlation, which will again allow us to make better predictions. Not to say that we can't make predictions today, and that's a false argument. We can make predictions today for a subset of variants and a subset of situations, and we have to know which or which. We do have to address the sensitivity problem. I told you that we're trying to set our threshold very high for the calls that we make. That leads us then to make a different error, which is that when we don't return a result to a patient, they may still have a risk for some of these diseases. And we do worry about patients misinterpreting the absence of a finding as a clean bill of genetic health and making incorrect medical decisions on that basis. And we work hard with our participants to do that, and I think as a field, we're going to have to struggle with that so that we say, yes, this is an extremely powerful technology when we find something, and when we don't find something, it's not a strong negative predictor of risk, and that's a hard message to get across. Changing the paradigm which we're doing causes discomfort, and I talked to you a little bit about how it causes discomfort amongst our colleagues also causes discomfort amongst ourselves, and I think it's within our own field. We have a lot of challenges, and that leads me to the college statements. The college has taken these sorts of scenarios and data into account and made recommendations for how incidental findings should begin to be approached. The statements actually comprise four documents that need to be viewed as a whole, and I would encourage you to look at all of those. I can circulate these if anybody needs them. The college comprised a working group of about 14 or 15 of us co-chaired by Robert Green and myself to actually ask questions about how we should approach the problem. The charge from the college was whether or not the college should have a policy and or a list of incidental findings that should be sought in the process of doing exome and genome sequencing clinically, only in the clinical realm. And if so, what that list should look like and what it should include as far as genes and phenotypes. What we settled on as a working group was that, yes, we should go forward with such a recommendation and we set some general parameters for what the traits ought to comprise high penetrance and or clinical confirmatory testing that would allow the prediction to be affirmed to be true. Long asymptomatic period meaning that there was opportunity for misdiagnoses that the sequencing could correct. Some kind of efficacious treatment screening avoidance process, et cetera, that would markedly increase the patient's health or survival disorders not detected by newborn screening. We focused on newborn screening as a separate realm. And then we focused only on a subset of highly pathogenic variants and agreed for a lot of reasons that this list was going to have to be updated constantly over time going forward. So what we set out was a minimum list of medically relevant or you could use the word actionable conditions and that the variants and the genes that we specified should be returned to clinicians and this is the key thing. The recommendation is that the variants should be returned to the ordering clinician is what we say and then the clinician has to contextualize that and determine when and how that result is returned to the patient. We controversially did say that the variants should be returned regardless of the age of the patient which seemed to some to contradict prior guidance on this topic. I'll get to that in a minute. And we estimate from some of the work that we and others have done that in the order of about one percent of people who undergo clinical exome or genome sequencing would get a result from the list that we generated. Now it is absolutely clear that these recommendations diverge from current clinical genetics practice. No doubt about that and I think that's because the change in the technology mandates that we change how we handle the patients. What's interesting though is most of the recommendations actually converge with practice in other disciplines of medicine and I'll talk about that more in a minute. Controversies, preferences. We, this recommendations clearly require informed consent and that's a big part of the recommendations. What they don't include is preference setting by patients. We did not settle on a notion that patients would get this pick and choose amongst the disorders to decide which of them that they want and that overturns a long held practice in genetic medicine which is the patient decides which tests they undergo. What we reasoned is that the incidental findings are an inextricable part of the clinical assay i.e. the genome or exome sequencing and it is a package just like radiology is a package where you get the indicated finding and the incidental findings whether you wish to have them or not. So again this overturned the precedence in genetics. And the pediatric recommendation is based on the transient nature of where we are in the field right now which is that genome and exome sequencing is not widely available. It is commonly done in a setting of sequencing a trio of individuals where usually it's the child who sequenced for a pediatric onset disorder and the parents are sampled for segregation or genoval status, etc. And since parents don't have already access to genome or exome sequencing today we reasoned that finding a variant in a child even though it's not actionable in childhood but it's clear medical implications for one or the other of their parents in almost all circumstances and until genome or exome sequencing is widely available we felt that because of that reason if you find an adult onset disease variant in a child it should be reported back to the parents because of its potential action ability in the parents. Again this is controversial. Physician thank you very much Pearl. Yeah you are, I appreciate that. Again so consent is a big part of this. There's been some unjustified criticisms that we don't endorse consent. This is exactly what the documents say. I don't think you could have a more robust consent policy and in fact it exceeds the consent policy that we have in all other areas of medicine for all other incidental findings. So arguably I think our consent policy is too strong when you compare us to the rest of the medicine. Other clear controversial element is the level of evidence and it is absolutely clear that there isn't sufficient evidence on an objective randomized controlled trial basis of how you could do these studies to justify everything that we recommend. These recommendations are based on expert opinion of a process that I would argue is very thorough. This process took about 14 months of committee calls. A lot of input from membership of the society and the board of directors including an open forum on the column. And Robert Green pointed out this statement to me. The U.S. Preventive Services Task Forces sometimes use it as a cudgel to abuse people who don't have robust evidence but I thought it was really interesting he found this quote from Preventive Services Task Forces. Even though the evidence is insufficient the clinician must still provide advice. Patients must make choices and policy makers must establish policies. This is not a reason not to have evidence. We all want more evidence but we also have to acknowledge that we live in a real world where this technology is being ordered. Thousands of genome and exome sequences have been already performed clinically and we felt that it was appropriate to use the best knowledge that we have today and make a sensible statement, acknowledge that the data are imperfect, that it will evolve over time and that we will improve these practices and avoid the chaos. So making these recommendations has been a really good learning exercise and some of you might recognize this is a device called a Rorschach block and I think these recommendations are somewhat of a Rorschach block for our field and they're telling us a lot about how people view medical genetics, medical genomics and predictive medicine and in some cases, I don't want to get too psychoanalytic on you here, I think they're actually telling us more about the people who are criticizing these recommendations than they are telling us about the recommendations themselves and what people bring to this debate are a lot of long held and cherished assumptions that seem to my mind not to recognize the fact that the technological landscape has changed underneath us. John Maynard Keynes said in a rebuttal to a criticism he once got, he said, well sir, when my information changes, I change my conclusions. What sir, do you do? And the truth is some people don't change their conclusions when the facts change and I think that's sort of where we are in this debate and we're learning a lot about that. Some of these concerns that have been raised are absolutely valid and I think in some ways we as a field are afraid of ourselves and our shadow, if you will, is a legacy where we have this terrible history of things like eugenics, we have a long history of diagnosing disorders for which there is no effective medical treatment. We also have a long history. The most established part of genetic practice is in fact in prenatal diagnosis where we have no business making recommendations to patients about what they should or shouldn't do with those data. So those events have conspired to make us be extremely reticent as a field to use a piece of genomic data and tell a patient you should do this. But now again, the landscape has changed and we have to recognize there are some situations we shouldn't always be afraid of our legacy and our shadow and we should actually become comfortable with saying to patients, here is what the deal is, you need to do this to keep yourself alive and healthy and we have to change that culture and I think the challenge is more in ourselves sometimes than our colleagues. So I'll stop there. I've certainly enjoyed being part of this process. We have been the target of a fair amount of criticism but I think that is a good thing for our field that shows that this process is really important not only to us but to the entire field of medicine and we're changing the terms of the debate and moving genomics into medicine and we all have to acknowledge that's going to be hard and we're going to get a lot of criticism when we try and do that but it's worth it nonetheless. I'll stop there. Take any questions. Thanks. So we're running late but I think we'll take a couple of questions. Yes. Deborah? I think it's very interesting in your ClinSeq program you did the phenotyping yet the ACMG recommendation is that you return this result to the clinician. When do we start being the clinicians who are specialists in what the genomic information means because this result is going to go back to primary care physicians who don't necessarily know what to do with this information and I think we should be stepping up to the plate to work with patients directly to have them understand what this means and then refer them to a specialist. I mean when you send someone to an oncologist they don't send the diagnosis back to the primary care physician and say so you give the chemotherapy and I think we have to start thinking about ourselves differently. So I think when we're in the realm of highly penetrant disorders that are Mendelian or nearly so I think the medical geneticist is a person who has a lot of expertise in these kinds of questions and it is the case in all realms of medicine that it is a dicey proposition when a clinician orders a test when that clinician doesn't understand how to interpret and use the data. Now that's not to say that they can't order because we all know they do but the recommendations are quite clear in the guidelines promoted by the college and Bruce you might want to comment on this are quite clear that the person ordering the test should have the expertise to interpret the test and there is a good reason for that. So did you define who should order? That's so much genome sequencing because I think that's a very good question. It is a good question. My personal view is that the college has a slightly different policy than my own personal view so I'll give you mine first and then I would ask Bruce to comment. My view is that I wouldn't say that it has to be ordered by a board-certified medical geneticist nor would I say that the counseling that is also recommended to be part of this process has to be provided by a board-certified genetic counselor. I think that's a little too narrow and unrealistic of a scope. I think what we should strive for is that the person ordering the test and talking to the patient understands the core of what the test is about what the implications are for the patient so they understand it and then upon receipt of the return of the results can manage them and put them in the correct context for the patient and get them to the right person when they have a result because the clinical geneticist should not be implementing the colonoscopies for the MSH2, right? That's silly and they need to have a network of providers and that's how all of medicine works is we have to understand what our boundaries are go as far as we are capable and comfortable going and not go farther and who that person is neurologists are going to order a lot of exams. They can learn how to do this. I'm absolutely convinced they can. Bruce, do you want to comment on that? So I don't think this is that different from a lot of areas of medicine. I'm licensed to perform surgery I think probably somewhere along the way but I don't think it would be very wise for me to do that and I think there are many examples, the example of not pushing chemotherapy if you're not trained in that is another such example and there's no law that says you can't. It's just that you know that you're likely to get into trouble. So part of what we have to accomplish here I think is to educate the community about what the power can be and where things are applicable in daily practice and where it is highly advisable to seek the assistance of people whose training is such that they would be better qualified to do this and I think this is an example where that is the case. Now the college in its points to consider does make a strong statement about the need for a person who is trained in genetics and genetic counseling to provide both the pre-test counseling and the post-test counseling and doesn't go to the extent of saying it has to be a board-certified medical geneticist although it does recognize that there are such individuals and as far as I'm concerned this is why there are such individuals just there's a reason why there are oncologists who are specially trained and so it seems peculiar to me to talk about creating new specialties or repurposing specialties when there is a specialty that is qualified to do this and has the experience. Now you could argue there aren't enough of them but that's because this is a changing field and until recently probably the potential of applying this on a very wide population scale didn't exist. So it may create a pipeline problem and I can tell you the college is actually working now to try to overcome that but that's not an argument not to do it. It's an argument why a new paradigm exists and there's a lot of opportunity for people to enter an exciting field. So I just think this is very much like other areas of medicine. It's an easy box to check perhaps but it is not an easy thing to do and requires some special skills and training. I think we've really got to move on because we're running quite late but this is clearly a topic that there will be a great deal of additional discussion about. So I think next, well then maybe you can use that as a starting point to expand this discussion then for a little bit of time.