 differentially methylated genes that we identified. And interestingly enough, first of all, the most significant ones actually are involved somehow, many of them in P53 function, which we haven't pursued yet, but it does suggest that these genes are in themselves important. It's just not patterns that we're looking for. But this is the first particularly interesting thing. If you look at the difference in this heat map, the methylation pattern of these genes is entirely different if an individual carries a mutation in P53 versus if they do not. And that is because the bulk of the patients who carried mutations were in this cluster over here, there were very few of them over in this cluster that do not show differential methylation. This is a blood test. So these samples are being tested in the peripheral blood lymphocytes of patients. We also have tumors for a smaller subset, and the methylation patterns there are also somewhat different. The second part of the observation, which we're now really want to pursue in looking for much larger data sets, and this is where we've been having some offline discussions with colleagues here about sharing our resources back and forth so we can do methylation sequencing on larger numbers, is here. Again, you see a difference in pattern, and from the back of the room, you might say, well, that's kind of hard to know what I'm looking at, except there you see a number of individuals with a little green dot. Those are patients who have adrenal cortical carcinomas, and virtually all of them cluster differently than, for example, patients who have brain tumors, which is there. This suggests that there is actually a signature as measured in the blood, in peripheral blood lymphocytes, which actually has some predictive value to tumor type in an individual patient. This has nothing, well, I shouldn't say it has nothing to do with the P53 mutation, but for some of these, there are different family members with the same P53 mutation who develop different tumors and fall into the different categories. So methylation, we believe, plays a very important role in defining phenotype. This would be incredibly important in terms of predictability and predictive testing. And then we started to dig down into what some of these genes were, and just out of interest, there is a microRNA, Mir34A, which is the most significantly differentially methylated. It's hypo-methylated in P53 mutation carriers compared to normals, and it turns out that this particular microRNA is actually very important in regulating P53 function, so it's no surprise that its dis-expression or abnormal expression in the context of a person who carries a mutation may have something to do with the development of cancer, okay? So as we've seen and heard and referred, the concept that there is modified genetic events that are important in determining what type of cancer and when is clearly important, and here are just some of many, these are published studies or others not published that we're all quite learning more about that will probably tell us a great deal that we could use as a predictive test. And one could imagine maybe in the future that a patient with a mutation in P53 will be able to have all of these tested, have almost like on a little card, and that can be put through an algorithm to determine their individual risk of developing a particular type of cancer. Okay, so I know people here want to hear about the Toronto Protocol from somebody who lives in Toronto, no offence Olivier, no, yes. But it's actually quite an interesting piece. So you've seen this slide before and we're actually in the process of making modifications based on the work that Christian had led from the American Association for Cancer Research of combining this international effort, some modifications. One of the modifications to the pediatric screening which is here is I'm gonna be pushing very hard that a dermatology exam is done on every child starting from the minute that we know that they carry mutation. We now have two children in our practice who develop melanoma, one of them is 13 and the other is nine, okay? So when the first one developed it, I asked all of my colleagues what the youngest that they had seen and it was above 20 years of age, we know it can occur, so we have to be doing at least annual dermatology exams. The rest of the screening were probably sticking with pretty much similar to what we have for the time being, but we'll see how that evolves. And this is data which is published, but basically to show you the survival curve here which you've already seen that Olivier had put up. And I think it is true, I think that the lessons that we're learning and everybody else's learning is that we need more numbers, more patients, more surveillance and we also have to teach our radiologists in particular as much as possible. They need to teach us what the best ways to develop technologies are so that they can be as accurate as possible on the picking up of lesions and whether they are true positives or false positives or false negatives. So, but an example, since the study was done, we've had a number of newly diagnosed patients based on screening and this is just a list of some of them here. You can see the third one is one of the patients with 13 with the malignant melanoma. And here are three examples. There is at the top a boy who presented with a small chondrosarcoma of the rib. It required resection only and he's now a year and a half since with no evidence of recurrence. This one here is a 16 year old girl who's actually had seven malignancies before her age of 16 and on routine surveillance she was found to have a myeloid sarcoma in the abdomen but there's surveillance actually picked up a breast lesion and that's a phyloides carcinoma of the breast. She actually two months ago had a bilateral mastectomy with reconstruction and we're continuing to follow her. And then the third one just below is a young girl completely, all of these were completely asymptomatic who had a diffuse astrocytoma grade two resected and she's now about four years post that resection and so far doing well but we're watching her as well. And so these are the kinds of things that we are able to pick up with the surveillance. But here's I think a very important piece is that with the surveillance regardless of whether it's by MRI or ultrasound or physical exam or lab work there's three categories of things that you will tumors that are found as you've heard benign, pre-malignant and malignant. The ones that we worry about that we've always worried about are the malignant ones but we would make the argument that the pre-malignant ones we also have to be concerned about because they have not acquired yet the other somatic genetic events to completely transform them and make them technically more difficult to treat. The advantage of detecting things as pre-malignant lesions whether it has a long-term effect on survival and I think it's fair to say that the life screen suggests that perhaps it doesn't at least in the shorter term but what it does do is it reduces their morbidity because the therapy is much less. None of the patients in that middle group receive chemotherapy or radiation. They all were treated with surgery alone. So this is an important piece but the second thing is it turns out that with surveillance about two thirds of the patients tumors that were identified were actually pre-malignant. In other words you could treat them with effectively with minimal therapy whereas none of the patients tumors that were detected just by non-surveillance came with clinical findings were pre-malignant. They were all malignant. So there is a difference in the types of tumors we think that are coming out of these surveillance protocols. And you've seen this was the follow up of the larger scale study. Okay, so Dr. Lee who was one of my mentors had said that in essentially from many syndrome is an experiment in nature and Fred was an incredible clinician dearly loved by his families and patients who he worked with. And so this wasn't meant to be that patients were guinea pigs, absolutely not. What this was is it's telling us something about nature that we need to understand and we can learn a great deal from our patients to try to explain that. And the example that I give here from the surveillance is kind of interesting and does tell us something. So this is a young girl who presented with an anoplasmic Radomar sarcoma in 2001 in the McZilla, so where the yellow arrow is. You can skip this second one. And then 16 years later, almost she presented again with the yellow arrow with an osteosarcoma in the radiation field just on the edge of the radiation field. This was treated with surgery alone. Actually, I'm sorry. That was surgery, chemo and radiation. This was surgery and chemotherapy because the surgeon we weren't confident could get the entire thing out. And she's now two years after treatment and looks amazing and has done really very well. But what is interesting is if you look in this middle scan and I'm not a radiologist and my radiology colleagues tell me that in fact about 200 days before at one of the earlier screening scans, there actually was a lesion there. It just wasn't seen. It was a false negative in retrospect, okay? And so this is a lesson. But an even more striking lesson is here. Young boy with a chondrosarcoma at the age of 18, all right? Now, it turns out that he was lost to follow-up. He's the only patient that did not come back for his annual visit. And the reason was that his father in the interim had passed away and he really, it was a single father and three of his siblings had already died of cancer. And I think that the psychological devastation of this kept him away from hospital. But he did eventually come back. We called him and called him and eventually he came back and he presented with that very large tumor which actually has been resected and he's been fine. But in retrospect, going back 900 days, three years, there is a little spot on the MRI which our radiologist or convinced was actually a lesion that if we had known at the time or seen it or reported it, probably would have been resected or at least biopsied. So if that had happened, it doesn't take a radiologist to tell us that that would have required a much less minor surgery than what he eventually had. But it also tells us that cancer can remain dormant for many, many months if not years. So the P53 mutation is initiating in that event but there's many other things going along that we have to understand before the cancer actually presents clinically or that we pick it up by surveillance in an effective way. And so that's the experiment in nature. Can we understand that? To try to do that, you can't do this experiment in humans because you're never going to leave something that you know to be cancer and watch it grow but you can do it in mice. And so we've got P53 mutant mice that we now have about 70 of them where we do monthly MRIs. We put them to sleep, we wake them up. We also draw blood. I say we, I don't do any of this. My students do it and they're amazing. But Sangeetha who does this draws blood from the tail fade every two weeks for CTD analysis. But the other thing we can do with the mice is this. When we see the tiny lesion on the MRI we can actually resect or biopsy it and then close the mouse up and let the mouse continue to grow and see how the natural history of this lesion develops. And so we can follow them in this way to determine how we're doing. So in the last couple of minutes I'm just going to go to where we are now. So what we're trying to do is develop better predictors even better than any of this surveillance. So working with a computer scientist and a Goldenberg in our group. What we were interested is taking all that methylation data and seeing if we can use it in any way to predict anything. And the first question was can we predict age of onset? So I don't really honestly understand all the mathematics of this. I was a physics major when I was an undergrad, not a math major and for sure not a computer science person. But algorithms are developed to be able to model predictability based on certain variables. In this case taking the variable of certain age of onset. What we say it is can you predict the age of onset just using the methylation array data? And using that we were able to say can you predict before the age of four whether the first child will develop cancer or not? And you can do so with a false negative rate of about 12%. It actually gets worse when you try to do it at about five years of age but it actually improves when you're around six years of age. So there's an 89% likelihood that you will pick up the cancer if it's there and there's an 11% likelihood you'll miss it if it's there. That's actually a very good sensitivity rate for most clinical tests. But what's also important that we've heard a little bit about this morning is by the age of six, most kids can tolerate an MRI scan without any sedation. So if we're able to predict that the child will not develop cancer before the age of six using a test like this, we wouldn't bother with any of the surveillance. They don't require sedation. We'd only start after the age of six. However, if we have high confidence that they will develop cancer before six then we will have high confidence that notwithstanding the potential problems with MRI sedation gadolinium, it is still worthwhile because the benefit of identifying cancer far outweighs any of the theoretical risks that we yet know about early sedation. So this is actually quite exciting. Now what in the mix of the computers, three days ago we got another piece of data off the methylation, which I think is even more interesting in that what we've done now is taken the methylation data, laid it over the whole genome sequencing data to look at copy number variation, single nucleotide variation and methylation. And when you do a multi-level multiple factors in this way, we actually are now looking at in the first 35 patients that we've done, regardless of age and regardless of gender and regardless of tumor type at the moment, you can see that you can predict actually who does get cancer from who doesn't get cancer. And many of the ones in the doesn't get cancer, they're adults and they're there in their 30s or 40s already. So it's interesting. So this would be also a very valuable test to have that you can actually determine who may or may not get cancer. At the end of the day, you still need blood to do these. These are blood tests. And so the last piece then is going back to that first challenge. What is the molecular basis or signature or architecture of a tumor from a patient with leaf from any syndrome? Is there something genetically unique about those tumors that you could actually detect in a blood test? Circulating tumor DNA, as Dr. Caron has mentioned earlier. And I think the answer is probably going to be yes, but we're not quite there yet. And the two clues are this. In one study by this student, Nick Leith, what he did is he's taking multiple tumors of different patients, cutting them up into teeny, teeny pieces and doing sequencing. And doing that kind of sequencing, he's actually, I sort of need an arrow for this, but if you just look at the bottom, there are signatures of genetic alterations that are unique to different types of cancer or chemotherapy exposure, UV exposure, things like that. And in the bottom, we actually have a number of signatures that appear to be common to all the patients who have a P53 mutation germline and different kinds of tumors. And so what it would suggest is we don't know yet exactly what the signature is, but there is a pattern that's consistent across all of these. If this signature is small enough genetically, in other words, we can't go and sequence three billion bases for every single patient. But if it was 20, 30, 40, even 100 or 200 bases or 10 genes or 12 genes, something manageable, then you could do that at a clinical level. And the reason I believe that will happen came from a study that was done, and we had nothing to do with Lee-Franmeni syndrome, is a recent study from a group in Toronto actually studying leukemia, myeloid leukemia, has identified a 21 gene signature that was predictive of the development of myeloid leukemia in adults. And it's a simple test, right? There's one for breast cancer, there's others as well. So I think that's where we want to go. And to sort of skip forward, what we've done is we've linked the genomic data that Nick's doing, as I mentioned, with the methylation data that Valley, one of the other students is doing, and come up with a signature that's a methylation signature that can be picked up off of the blood. And what this shows is a plot where the green, our patients with Lee-Franmeni all have a P53 mutation. There's only 10, it was the first test pilot of this study. And then all the other dots are patients with all sorts of other cancers. And if you look carefully, you can see that blue tends to cluster with blue, red with red, black at the bottom with black. So different cancers tend to have methylation patterns that are similar to each other, even when you measure it in the blood. These are blood samples. But the Lee-Franmeni patients samples separate completely separately, okay? They're sitting on their own. So they are different than people who get cancer who don't have Lee-Franmeni syndrome. And when you look at it in a different way, it turns out that in fact all the patients, whether they have cancer or not, the 10 at the top there, have a different pattern of methylation that actually we believe will be able to also be predictive of the tumor type that they would get. So the combination of methylation plus sequencing may narrow our ability for early detection. And then we could use that to guide the kind of clinical surveillance to try to find when the tumor actually starts. So instead of every patient getting all the surveillance, you can actually, as Pierre has suggested, that there's this sort of areas or ages or time periods when people are more or less at risk, now you may be able to narrow it down to time periods and tumor type. And that's what we would like to be able to do, okay? So in summary then, what I haven't talked about is some exciting work that's going on in many areas for therapy. That would be a whole different lecture, but it is, I think there's a lot of hope. This is really speaking to the families in particular, who not only do we wanna prevent, but for those who develop cancer, sorry, do we wanna detect, but are there ways that we could prevent the cancer from developing? And I think I can say that over the next few years, not tomorrow, not next month, probably not next year, clinical trials for a multiple types of molecules that are able to manipulate P53 so that it is behaving normally are out there and being tested. And with time, those will find their way into the clinic so that the concept of chemo prevention will override the need to do any of these kinds of detections. But while we're not there yet, the hope is that we have a model like this where multiple genetic events are occurring in an individual who carries a mutation, some catastrophic event occurs in the particular cell which leads to the cancer and the final iteration, but some form of surveillance is able to block it at that transition point between a benign or a low grade tumor to a malignant one and you sort of can reverse the process. And as Thierry uses that very nice phrase of not really preventing the cancer, but not preventive medicine, but disruptive medicine, disruptive medicine. All right, this is the inspiration for my work as with you. These are teenagers from the Toronto, the Fraumeni meeting. These are all kids who have P53 mutation carriers, some with some without cancer. And they asked for a picture with Dr. Lane on the far right, Dr. Levine, the other co-discoverer and Joe Fraumeni over here. And I insisted that Joe sit beside me because I wanted to. And the person of course that we miss in this picture very much is Fred Lee because it was really the two of them who started this. That's my lab. And yes, we also have wine. It's not really the same and a wonderful group of collaborators and funders. Thank you. So many thanks.