>> Ravi Pandya: I’m very pleased to welcome here Chris Kemp and Carla Grandori. Chris has been working Cancer Research for twenty-five years. He’s a Full Member at the Fred Hutch Cancer Institute. Carla Grandori has also been working in the field for a very long, has been the Director of Sequencing Core. Is the researcher at both the Fred Hutch and at the University of Washington, as well as Founder of The Cure First Non-Profit Organization in the startup SEngine Precision Medicine. They’re going to both give a talk about some new work on cancer drug development and understanding. Chris you want to start. >> Chris Kemp: Thanks Ravi. Good morning, so we’re going to look at things from a different angle today. I’m going to give you first a history of the pro, a quick introduction to the problem. The status of where we are right now with sequencing and the promise of precision oncology. Where we think we can take this next. Here’s the problem right here. This was published in two thousand thirteen in the prestigious New England Journal of Medicine. Showing you that the survival of patients with pancreatic cancer is about twelve months and if you add this combination you improve it by two months. This is the problem that we face. Taking a step back to the nineteen ninety’s this is a very influential diagram, a very cartoon diagram of cancer going from normal epithelium of the colon to colon cancer. In this example going through the stages and this led to the dogma that cancer is a genetic disease. Here’s the mutations that are detected at various stages of cancer. I’m going to be talking about this gene is an aguA gene called K-ras mutated in about thirty percent of all human cancers. This gene is called P53. It’s a tumor suppressor gene mutated in half of all human cancers that ever exist or ever will exist. These are really major cancer genes here. This kind of spawned the Cancer Genome Project if you will. This is now updating you to two thousand fifteen. We now have these very comprehensive maps of the cancer genome of different cancers. I’m sure you’re familiar with this. This is an example from head and neck cancer. These are individual patients here. These are the genes that are mutated at certain frequencies in these tumors. I show this to illustrate two things. First the p fifty-thee gene that I just mentioned is mutated in about half of all, actually sixty percent of all head and neck cancers. This is the problem we now are beginning to get a glimpse of the underlying genetic features of these cancers. This is like a huge promise and hope. But the question is how do we deliver to these cancer patients? That’s really the theme of Carla’s and my presentation today. How do we turn this knowledge into useful information for the patient? Just to give you a picture of what P53 is. It’s a DNA binding protein that activates transcription of some genes. There’s the crystal structure of P53. To restate it’s mutated in half of all human cancers. There are no therapies despite thirty years of research on this protein. There are still no therapies to treat cancers with that mutation. Does the information that we get lead to a treatment in this example? Not yet is the answer. Here’s another example. This is pancreatic cancer. Again, here’s these genome maps of the genes mutated. These are the genes here at some frequency in pancreatic cancers. In this example the RasOCGO gene is mutated in a very high percentage. Something like eight-five percent of pancreatic cancers have a mutation in this OCGO gene and P53 as well at some potential frequency. Again, we have this beautiful data. This took a huge national effort of billions of dollars to generate the Cancer Genome Atlas which I think Microsoft is helping to analyze. The question is how do we turn this into therapies for the patient? That’s really the continuing question. This is what Ras is it’s a signaling protein. This is the crystal structure. It sits in the membrane of the cell and signals for the cell to divide continuously. Again, a third of all human cancers have a mutation in this gene. There are no therapies to treat Ras mutant cancers. The National Cancer Institute has started a fifty million dollar project to target Ras that we’re participating in. Realizing that again despite thirty years of research we really need to figure out and get serious, and all work together to figure out ways to target this OCGO gene. Here’s a picture I like to show which is kind of again if you look at all the mutations found in pancreatic cancers to date. KRAS is obviously very frequent, and P53 is another frequent one, and these other genes at other frequencies. We’re getting maps now. We’re starting to get maps. The question is how do we use these maps to get to therapies? This is a really, really major question. We read in the New York Times. We read in popular magazines about the cancer genome is going to promise all these future medicines. The question is where’s the answer? There’s a lot of hype around this field but where’s the answer? Here if you take a picture of a cell and the genes that are known to be mutated in cancers, and play roles in cancers. Those are highlighted in red. Here’s P53 over here. Here’s Ras over here. They regulate all these signaling pathways. This is kind of another map of the cancer cell if you will. They regulate things like gene expression, cell death, cell proliferation, etcetera, all the phena types that cancers have. Our approach and the major approach of drug companies, and I think one of the aims of the Cancer Genome Project is to find more of these and figure out how to drug those genes. How do you drug Ras? How do you drug P53? You’ll hear about MYC over here next from Carla. The answer is we can’t, yet. We still can’t so what do we do? Well, we propose and we now know that there are vulnerabilities in cancer cells that are not mutated genes. If we can target those vulnerabilities with drugs this may lead to effective therapies. I can give you one really, really clear and obvious example is breast cancer where the drug Tamoxifen is an antagonist of the estrogen receptor. The estrogen receptor in breast cancer is not mutated. Yet, it represents a great drug target. Tamoxifen has treated successfully many women with breast cancer. There’s an example of targeting a gene that’s not mutated. Sequencing the tumor would not have find estrogen receptor as a target. Yet, it is an excellent target. That’s one example and there’s many more. We’re going to find many more. How do we find those new genes? Like I said there’s a couple of examples. You know there’s estrogen receptor. There’s the PARP enzyme in BRCA mutant cancers. A few examples but this is I believe our current state of the map of cancer targets. It’s pretty old. This is like an old Portuguese sailor’s map. But we know we’re now convinced; we are absolutely convinced that there is a lot of land here, here. There’s going to be a lot of things to discover here. We’re just at the tip of this discovery process. I’m going to convince you today that there is indeed cancer targets out here. But we want to convert this old map into a very modern usable map for the doctor and the patient to use. To recommend certain therapies and to identify new drug targets for cancer treatment. How do we do this? We use a miraculous process called RNA interference. Here’s kind of a scheme of our process. We start with a cancer cell from either a mouse. We can isolate a tumor from a mouse or from a cancer patient. You’ll hear examples of that in a minute. We put them in culture. Then we can test every gene one at a time. We can do all the genes that are present in the genome. By making these RNA interference molecules these are just tiny synthesized inhibitory molecules. They’re placed in each of these wells. We put the cells in all these wells. We use; we can do these experiments in a plate of four hundred wells. If you do ten plates you’ve got four thousand genes which you can test. You put those; you infect the cells with these little siRNAs. You measure the phenotype of the cells in this computer based plate reader. You generate a rank list of all the genes that you just tested. These are the ones you’re interest in because when you put these little siRNAs into the cell you kill them. Now you’ve identified that gene as a target that you might want to target. You data analysis and this is a major, major sort of component that takes a lot of time and effort. We’re still perfecting this part of the analysis. Then you validate those genes as targets in the cells and in mouse models. That’s, anybody have any questions on this part because I… >>: What about the control of the non-cancer cell? >> Chris Kemp: Yes, we can, we do all kinds of match pairs of cells with or without the OCGO gene, with or without. You’ll hear a couple of normal cells as well, yes. Excellent question, yeah. Okay, so here’s an example where we can, of a larger library. We call this the Druggable Genome. There’s seven thousand genes in this RNA library because these genes are more likely to be the code for proteins which are more likely to be druggable. We have Kinases. We have Cell Surface Receptors and so forth. But this is just a list of all the genes that are in our Druggable Genome library. This is roughly; we have about thirty thousand genes in the genome. We’re testing a good proportion of those genes. This just kind of gives you a flavor of how broadly we’re opening our search. >>: [inaudible] siRNA. How deliverable is that? Here you’re working with usual cells kind of in culture can you actually target that siRNA in vivo [indiscernible]? >> Chris Kemp: In vivo not yet, as an actual drug, no not yet. But it’s a great hope in the future but people have tried and we haven’t got there yet. But it’s a great; many companies have come and gone with that promise. But people are still working on it. Actually delivering the RNA itself is a therapy. >>: Right, this is an asset drug therapy [inaudible]. >>: [indiscernible], sorry in your analytic pipeline this; the siRNA is just going to give you the knock down experiments. But in your analytic pipeline can you also try it per data function proteins that base off the knock down field what you can compensate? Or are you really just going for drugs that can take out of approach in this? >> Chris Kemp: Well, these genes up here when you knock these down the cells increase their growth rate. These are likely the opposite side of the coin if you will. In some cases these are tumor suppressors. These are OCGO genes. These are the genes that the cell’s required to survive. These are the genes that increase survival of the cells. >>: Those begin a function so you might try to enhance. >> Chris Kemp: Yeah, yeah. >>: Would you filter the RNA genes you’re doing the knockouts for ones [indiscernible] module drugs or… >> Chris Kemp: I’ll get to that. >>: Okay. >> Chris Kemp: I’ll get to that, great question. >>: You know and do you measure other things other than just survival like RNA expression and so on? >> Chris Kemp: Yes, you can measure… >>: Measure cells. >> Chris Kemp: Measure other phenotypes for which you can get a fancy microscope here. You could have that microscope look in every one of these wells. Look how the cell is behaving, the shape of the cell, and the DNA damage of the cell, other phenotypes if you will. That’s a big part of the equation and Carla… >> Carla Grandori: Yeah, yeah, we don’t measure routinely the knock down because these siRNA have been optimized. The secret behind this library was a library that was optimized by [indiscernible] formatics for over ten years with really optimizing the design of the siRNA. Now that library has been the algorithm for those siRNA has been sold to SIGMA so everyone can buy it. We do check it as a validation then we go back in. But in some ways it’s a library the hsiRNA has already been pretty validated. >>: I think maybe what Ravi was trying to get at is have we spent time thinking about leukemia’s? One of the possible therapeutic approaches to leukemia is rather than killing the cells to get the blast just to differentiate, which might be a phenotype that you can check. >> Chris Kemp: Yes. >>: It targets like cell death. >> Chris Kemp: Sure. >>: But you know you could check out surface markers. >> Chris Kemp: Yes. I want to emphasize that this experiment takes about a week or two weeks which is amazing. >>: Yeah. >> Chris Kemp: You could do seven thousand experiments in two weeks. >>: That’s seven, yeah. >> Chris Kemp: Yeah, that’s basically what we’re doing. Then you could also have many endpoints besides cell survival. You could have differentiation, migration, DNA damage, other endpoints you want to screen for. This is a whole new way of suddenly we have the ability to understand the cells function comprehensively in a two week experiment. It’s just, its mind blowing, simply mind blowing. Here’s, we’re going to jump right ahead now to. Let’s get right to the clinic here. We have a team at the Fred Hutch, as well as OHSU. We’re collaborating with a number of cancer centers around the country to really get this platform in the clinic. Here’s an example, from surgery a pancreatic cancer was isolated from a colleague, put into to culture. Here’s the pancreatic cancer cells’ growing at low passage. When you put these cells in mice they form tumors so that’s good. We’ve confirmed these are tumor cells. Here we divided the cells into two avenues of analysis. We did whole exome sequencing which is, you guys are very familiar with that which is what everyone else does. This patient had a mutation in KRAS and P53. Okay, not unexpected. Then the question becomes what do you do with that data? Well, what we did then we then in parallel did a druggable genome RNAi screen on this patient’s cells. We tested seven thousand siRNAs or seven thousand genes for their affect and viability. Here’s kind of a plot of those seven thousand experiments over here. We integrate and I’ll just go to the next slide. First there’s sequencing result, so it’s got a RAS mutation like I just said. It has a P53 mutation like I just said. >>: One P53 mutation. >> Chris Kemp: One P53 mutation. But it has one hundred and seventeen others. This is the kind of list you’re going to get. We just talked about this Ravi for your pipeline of picking these variants of unknown significance. Many of these have never been seen in pancreatic cancer before. The question are these drivers? What are these genes? How do you sift through this list? This is what everyone is getting. This is what Foundation Medicine is getting. >> Carla Grandori: Okay, but it’s good in some way, it’s good… >> Chris Kemp: It’s good information… >> Carla Grandori: Because some days we will have… >> Chris Kemp: It’s a start. >> Carla Grandori: Some days we will have that… >> Chris Kemp: It’s an anchor. It’s a start but we need to take it further. How do we take it further? Here’s an example of the patient that has the KRAS mutation. You go to Foundation One. This is the report you get. It is unlike, there’s no therapies that they can recommend with that, just to reemphasize the point we need to take this to the next step. Here we’re integrating, the red dots here are the genes that are mutate in this patient. The gray are all the genes that we tested, the six thousand genes we tested. Down here are three hundred and twenty candidate drug targets that when you knock these genes down you kill these pancreatic cancer cells from this patient. You can see there’s you know a small number of genes mutated, a dot is eight. But now we have a much greater number of potential targets. >>: Just a question, when you, there’s also just in fact [indiscernible] normal cells. It’s in differential… >> Chris Kemp: We’ll do that later. >>: [indiscernible]… >> Chris Kemp: We go through that and do that. >>: [indiscernible] >> Chris Kemp: Yes, we follow up with that. >>: Okay. >> Chris Kemp: That follows up on that guy’s question. >>: Right. >> Chris Kemp: We follow up on that and I’ll get to that in a second. Here we tested seven thousand genes. There’s eight hundred candidates and this is just showing you a Phi chart of the, of what I just said. How do we prioritize these hits? We compare to the cancer genome atlas. Are these genes mutated at some frequency indicating they probably play some role in cancer? Are they over expressed in cancer cells? Have they been seen in mice models? Is there known biology? Are they druggable, which was another question? Is there drugs available right today for that target that we could use right away? I’ll give you an example. This is like a strong bioinformatic component of our analysis that we’re still trying to optimize and simplify, and make it kind of a punch key approach. But right now it’s we have one computational biologist working eighty hours a week trying to get this going. Here’s the three hundred and twenty targets that we’re going to select based upon those criteria that I just mentioned. Here’s the whole druggable genome screen. I wanted to show you that these are all the dots, these are all genes we tested, and that’s viability over here. Here’s our set. We’re going to retest and confirm. Here we compare viability versus normal cells. This is representing the differential to normal cells. This is kind of how, one step of validation. Now we’re putting genes, those dots are now, we now have genes named, associated with those dots. These are all genes we tested. Now, I’ve just showed this slide to illustrate that here’s the gene name. Here’s its official kind of longer name. Then it’s the differential toxicity of the tumor cell versus normal cell. >>: [indiscernible] positive types of normal cell? You just pick one because some genes you might imagine… >> Chris Kemp: We would like to build a database of normal cells, many different kinds of normal cells. >>: [indiscernible] >>: [indiscernible] not going to matter to that [indiscernible] of a person. Right, it’s all patient dependent for the normal cells probably. You’re right library, okay. >> Chris Kemp: We want to do a lot of different normal cells. >>: Right. >> Chris Kemp: Get a database of what kills normal cells no matter what. Then what looks differential to tumor cells? What looks differential to pancreatic versus breast cancer? We have this vision which is, it’s just, it will happen just when of having that database, right. >> Carla Grandori: Yeah, but I think at one criteria that we use is also selectivity within a cancer type with a certain mutation. It’s not so much a difference between normal. But it’s really the difference with you know anchoring that to a sub-type of cancer. Because if one, we found, we already are finding genes that kill no matter what, all cancer types. Those are your typical you know checkpoint of those cycle genes. Those do emerge, so when one finds the selectivity for sub-types is where I think that the true, there is more hope for targeted therapies that could be less toxic. >>: Sure, yeah I was just concerned about toxicity, right. >> Carla Grandori: Yeah. >>: I mean you knock down oxidated cross correlation and everything dies... >> Carla Grandori: Of course, of course. >>: You know… >> Carla Grandori: There is, so in this case you had one normal pancreatic line. Then we have normal cells for various tissues. I like to use and I will show in this screen normally primary cells. >> Chris Kemp: Also in addition some of these genes when we go back to this here. When I talk about biology there, a significant number of genes have been knocked out in the mouse. The mouse has… >>: [indiscernible] the model for, yeah… >> Chris Kemp: We look at that too in the mouse is often normal that lacks this gene. We know the mouse; all the cells in a mouse can live without that gene. Those are also good priorities. Good questions though. >>: You also do network analysis on patients so you could downstream and upstream candidates? >> Chris Kemp: Yes, yes, yes. >>: That’s understandable. >> Chris Kemp: Yes, we do some of that. It’s a little, it’s hopeful but it could be better. It’s a great question. It’s helpful but could be better. This is just to show you that, so again the list of our favorite genes we’re going to narrow it down and narrow it down. >>: Just to understand [indiscernible] graph on the left is now the viability relative to… >> Chris Kemp: Actually… >>: [indiscernible] to normal or what… >> Chris Kemp: It’s actually not. But this over here is… >>: [indiscernible]… >> Chris Kemp: The tumor versus… >>: Control. >> Chris Kemp: Versus control. >>: [indiscernible] >> Chris Kemp: We’ve done that. I just haven’t shown you here. >>: Okay. >> Chris Kemp: But we compared… >> Carla Grandori: I think you have maybe in the next slide some of the different… >> Chris Kemp: I can show you in the… >> Carla Grandori: This is just the repeat of the twenty. >>: Okay. >> Chris Kemp: Right, this is the repeat. We’re narrowing down the pipeline. Here we actually these are the same genes right here that are now are the smaller set of genes. We looked at these because there are drugs that target these genes. For example Wee1 is a kinase and there is a drug that inhibits this kinase. Now here what we’re testing here is the effect of the drug. We’re going, making a major leap here from and RNAi to a drug. This is a major leap. >>: [indiscernible] the drug instead of… >> Chris Kemp: Same assay but with a drug. We do it in this high throughput more. Here we’re measuring the viability of an IC50 of, which means the drug is potent. We want anything down here is the potent drug. You don’t want things over here that are just not potent enough. This is the relative of the tumor cells to normal cells. That’s a ratio of one right there. We want things up here that preferentially kill the tumor cells versus normal cells. That’s getting I think to the question. >>: Right. >> Chris Kemp: What’s the top guy? It’s [indiscernible] Cisplatin, right over here. This is a drug that’s the standard of care for pancreatic cancer. It does kill tumor cells at preferentially to normal cells. But it’s actually a pretty terrible drug. It has, you know it doesn’t provide, it doesn’t prolong life that long. It’s just what they have right now. But we have all these other candidates now to go for. MK is right here. I’m going to talk about MK. This inhibitor Wee1 in the next slide. It’s right here. We got very interested in this kinase Wee1 because it’s a very potent drug made by Merck. There’s the drug bound to the kinase. Has a very good IC50 of five nanomolar. We went right into mice and put tumors in mice at day one and randomized the mice, and treated them with a vehicle or this drug MK which inhibits this Wee1 kinase and Cisplatin, and Cisplatin, plus the kinase. The tumors grew more slowly. We hit the target. We measured that. Here we went from an RNAi experiment to a target, to a drug. This experiment directly lead to a clinical trial at the Fred Hutchison Cancer Research Center which is starting right now. Patients are being recruited where they’re going to be treated with this inhibitor developed by Merck. It’s now made by AstraZeneca. >>: Is that, so is that targeted selecting? Is there screening that you would do from patients to… >> Chris Kemp: We’re working on that to get more precise. In fact that’s one of the purposes of this clinical trial is for Eddy Mendez the surgeon who’s driving this trial to optimize those markers. It’s effective; by the way this was a P53 mutated cancer. As I mentioned at the beginning we have no therapies for P53 mutated cancer. The Wee1 kinase which we’re targeting here is not mutated in this cancer. But it’s a good target because it preferentially kills P53 mutated cells. That’s kind of a key point. Yeah, so here’s an example of five different cancer types, different cancer cells. This is I just like this picture because these are all the genes that are preferentially effective in these different cancers. These in the middle are those that are in common to all cancers. This is kind of a nice way of sort of; the vulnerabilities are both shared between tumor types and cancer specific. These are the ones that we’re very interested. Or we could say pick on these that are associated with P53 for example. Or these that are associated with this mutation or that mutation. These are sort of showing you the private and shared hits. Now we can kind of build a map for vulnerabilities of the P53 mutated cancers versus these other mutations. This is I think near the end. We’re starting to build another map of, this is the RAS pathway. That’s the RAS gene right there. Some of our screen results are implicating these other genes as potential targets for cancers with RAS mutations. We’re looking at both P53 mutated cancers and RAS mutated cancers. We’re going to build this map out even further. This is a functional map now. I want to emphasize this isn’t a static descriptive map. >>: [indiscernible] again that understand the variables. Those arrows are indicating genes where the RNA, siRNA any knock down has a significant… >> Chris Kemp: Correct. >>: [indiscernible] kill them… >> Chris Kemp: Yes, that’s correct, correct. >>: Okay, cool. >> Chris Kemp: Right, right. >> Carla Grandori: Compared to normal. >> Chris Kemp: Compared to normal. We really, really want to build this out. We want to just, because just, you know as I, RAS is a, we’ve know about it for thirty years or more, forty years. There’s hundreds and thousands of papers published on this gene. We still can’t target it with a drug. Perhaps you know a combination of targeting these other pathways will, and we really, like I said we want to build this and do this over and over again and build a map of what targets for RAS mutated cancers we could effectively use. I know we’re going to find them, we will. I just gave you an example with Wee1. Our team is, this is an older picture but here’s Adam. I think you mentioned you work with Adam. Olga is the Computational Biologist who’s really critical. Eddy’s the Surgeon who’s running the clinical trial in head and neck cancer. VK’s a breast cancer doctor. It’s a little bit out of date. But we are trying to build a coalition. We use to call it the Seattle Project to Target Cancer. But it’s now extending far beyond Seattle. We have to rename this. But we have active collaborations with a lot of organizations. These are some of the stronger ones with Sage Bionetworks, with Cornell Institute for Precision Medicine in New York City, OHSU, and the ISB Institute for Systems Biology, University of Washington, and Cure 1st which Carla will introduce next. We’re part of a national effort to identify new cancer drug targets. That’s funded by the National Cancer Institute. These are the centers around the country, The Broad Institute, Columbia, Cold Spring Harbor. We participate in monthly phone calls with this national group which is analogous to the TCGA Group which you guys are familiar with. This is very exciting. We’re going to meet in April in a face-to-face. We share approaches and technologies. We think that this program has great potential to develop new cancer drug targets. This is the meeting in China where we ran into Jim Watson. Carla talked to him about this idea. This is about five years ago now. This is a quote that we got from Jim Watson, the father of the human genome and the discoverer of the DNA double helix. >>: Can I ask a question? >> Chris Kemp: Yeah. >>: [indiscernible]. This is great data. When you look at it you, what’s your biggest challenge right now? If you were to like get help what do you need to move the ball forward? >> Chris Kemp: I think Carla will show a picture of that. But we can answer it here too, is we need to just do more of these screens quite simply and get data crunching abilities, to crunch through this data and prioritize these targets. We need to get access to more samples from patients and reach out to doctors and patients. I’d say that’s… >>: [indiscernible]… >> Chris Kemp: Technically, technically, we don’t have any technical challenges in front of us. We figured them all out. We just needed to push the button and build the coalition that I talked about, really. >>: Well the private and public industries, and companies you would… >> Chris Kemp: Yeah, we’re working very hard on building those ties with, yes philanthropy, for profits, academics, whatever it takes. >>: You think, can people get a hold of you at Fred Hutch? Or what’s the best way to… >> Chris Kemp: Yeah, they could get a hold of my email at Fred Hutch and Carla will introduce her, she has a different organization that she’ll talk about. There’s websites. There’s a Cure 1st website you could go to and Fred Hutch website you can find me there also. Numbers of samples we’ve done is we’ve screened about I’d say fifty or sixty tumor cells. Of those about I’d say ten are patient derived. The patient derived aspect of this is very novel. You know anyone can take a tumor cell from a, you can buy them and do these, do some screens on them. But the novelty of what we’re doing is we’re actually screening the patient derived cells directly. We can do this within a short period of time from the biopsy. Like a period of a week or month somewhere in that, between a week and a month, and have a report for that patient. Carla will talk a lot about that. We’re just, like I said there’s no technical hurdles, there’s just simply ramping this up. I think you’ll show your vision of how that could look. If we don’t answer your questions please come back and ask them again at the end. Okay. >> Carla Grandori: Yes, answering your question David what do we need? We have too much data we need people to help us integrate with the genomics. I think I feel that one thing I don’t want to underestimate the incredibly importance of the TCGA Project. I think having that genomics data is amazing. It will certainly pinpoint to new drug targets. It would allow us to integrate this functional data and anchor to you know DNA changes that we can actually measure very accurately. Have biomarkers so that a patient can be directed to proper therapies according to those. That’s why we’re here. We’re here to talk to the computational scientists of Microsoft. Think about what we could do together. We’re really next door. We all share this beautiful city. I think we have a mission to really bring a personalized treatment back to patients. That’s why we’re doing this. We have two organizations, Cure 1st is a not for profit organization. It’s, in fact we are on the list of the Microsoft endorsed charities. We founded, Chris and I founded it about five years ago with the help of really local people, doctors, intellectuals, writers that have helped us to set it up. The idea is to fund high throughput screening projects and advance this really exploration of the world. What we now call a galaxy of potential targets. Now SEngine Precision Medicine instead has started with the idea of what can we do for patients right now? Is to take a sample and I will show, and really test with all the possible drugs and match patients with drugs. That’s the timeline. I already went through here showing that how first was Cure 1st really driven by patients amazingly. I started the Quellos High Throughput Screening Core in two thousand and nine at the University of Washington, where I was given. Thanks to a philanthropic donation I was given a large room, empty, and said we need to build high throughput screening. Here are three million dollars, go. I went and built this capacity that serves the University of Washington, serves you know across the country. People bring samples and screens are performed there. We still do the SRNA screens there at that facility. Meanwhile I, again, while I was director of this facility Oliver a patient with melanoma contacted me to have, to ask me if I could test his cells. At the time I never thought of that possibility. I was very focused on the research you know using cell lines. I thought we can’t do this. Thanks to Oliver we began thinking about actually doing this for patients. This has actually happened thanks to the grant that Chris led in two thousand fourteen we were awarded this grant from the NCI. Then later on through Angel investor we started SEngine Precision Medicine. It was really the convergence of technology and also an intellectual if you want concept. That brought us to think about approaching identifying new way to treat cancer. It was through the concept of synthetic lethality. This concept was brought forward by Lee Hartwell that was the former director of the Fred Hutch. We can’t claim that as our concept but we applied. Thanks to the use of robotics, RNA interference and having the human genome sequence we could go full speed to identify new drug targets. This is, I think you might be familiar with the concept. The idea, that a gene is essential only in the context of cancer cells but not to a normal cell. There’s one other example that is in the clinic is for example the use of PARP inhibitors for patients that have BRCA one mutation. This is feasible and our approach is the arrayed approach versus the pooled lentiviral approach. You may have heard this at the Achilles Project from the Brody Institute. They use the pooled approach. Now in the pooled approach you mix all these lentivirals in a pool. Then you ask what drops out? What’s selected against? In general negative selection screens come with a lot of problems, low sensitivity. Imagine in fishing you know the needle in the haystack. I always like the idea of the arrayed approach. We are measuring precisely into each well the effect of each gene knock down. To a comparism for example, we analyze three thousand genes. We arrive with about fifty hits that we confirm with you know ninety-eight percent confirmation. Screen done at Harvard, thirty thousand genes, this sounds really great in, you know when one presents is that our, we can look at thirty thousand genes. The reality is that the number of high confidence hits is very low because of the, again the inherent problem with pooled lentiviral screens. The pipeline for siRNA screens and data processing has been established. Much thanks to Olga Nikolova and in Chris’ lab. Here we show that it producibility of our screens and various use of replicates. Here I want to get to one of the problems that I’ve tackled using the technology and it is MYC. Chris talked about P53, KRAS, and MYC is third in the run for high incident. In many cancer types, ovarian it’s about forty percent amplification. MYC is a transcription factor very hard to drug. We setup an isogenic screening in an isogenic setting. We isolate normal fibroblasts that prolific very well from circumcision from Swedish hospital. Those are what we use as our control. The other interesting part is that we can put OGCO genes and in parallel observe the affect of that OGCO gene. These cells will withstand the affect OGCO genes. Most cells will undergo [indiscernible]. We can do it for any gene that’s mutated. It’s a great system to analyze cancer specific vulnerability. This is what the output of a screen saw normal cells versus MYC overexpressing cells. We can see that if the genes fall out at diagonal these indicate that the gene preferentially are requiring in the context of MYC. This is what it looks in reality. I will go quickly because the data are published in two thousand and twelve. Again, we had about fifty genes that we identified. Obviously we want to find genes that are important in cancer context rather than into normal cells. We interrogate a neuroblastoma because it’s a disease very clearly driven by MYC. It’s a pediatric cancer. It’s a tumor that originates frequently in the adrenal gland. There are cancer types that have MYC amplifications versus others that don’t. You can see even though right now it’s all treated as high risk disease in high stages. The non-amplified and the MYC amplified are two really different diseases from the functional genomics point of view. There are target for the MYCN amplified and target for the non-amplified. We integrated the results from the isogenic setting into neuroblastoma. We went also one step further here we interrogated seven thousand genes. We built a network of genes and in this case are the protein encoded by this gene and how they interact. Can we really find a functional group? This also has been published now in the Cold Spring Harbor Perspectives in Medicine. It tells us that if we had to tackle MYC we can probably tackle by going through the DNA, certain DNA damage checkpoint. We can tackle a lot of the transcription on machinery. Now, at the time we did the screen we felt well that’s going to be very difficult to drug. But now we know BRD4 is a bromodomain containing protein. There are several drug companies now that have been able to develop drugs. In fact, this pocket of these proteins that allows to target drugs are the derivatives of Benzodiazepines which are you know tranquilizer molecules. It’s a highly, who would have thought that these type of molecules could be drugs for cancers. I think we’re going to have more of those examples. We also found that if we inhibit the inhibitors of MYC that was a surprise. We also can push the cells over to cell death. This is a paradox that has been seen and it’s because being an excessive amount of MYC if one goes even further and increases the level the cells cannot tolerate. I would not choose that as a target. But it has been seen in these overexpressing screens, also in yeast, and in other systems. We wanted to see what’s common between neuroblastoma and ovarian cancer both with MYC amplification. We found six genes in common that functionally inhibited the growth of ovarian cancer with high MYC levels and neuroblastoma. These are the genes. We are hoping to develop drugs against some of these genes. This is part of the reasons to start SEngine Precision Medicine. We also integrated the results with TCGA. This is the preliminary analysis by Brady Bernard at the ISB. That shows the empirical and the TCGA integration points to similar genes. CK1 Epsilon, PES1, two genes that are cross-cancer types correlate with MYC. Interestingly one of our favorite targets that emerged from TCGA and empirical testing is CK1 Epsilon. At the time we discovered there were very few papers on cancer. Most papers concern regulation of Circadian rhythms. In fact, both a small company in Seattle and Pfizer had developed inhibitors toward CK1 Epsilon. The inhibitors were designed to cross the blood-brain barrier and alter the sleep pattern. We tested these inhibitors in my, both with neuroblastoma and in ovarian cancer and show that they were effective at inhibiting the growth of the tumors. We’re hoping now to, this, we’re done with the first generation of inhibitor. We’re now planning to test the second more refined molecules. I always bring this example that I would be the first one to try this drug. Because I’ve seen the effect that this drug has on to normal cells, it hardly touches them. If I have to sleep less for a couple of days but you know I would be willing to try that. I realize that this talk is registered, oh well. [laughter] I should be more careful. But anyway, so optimal, I will end on the potential of the identifying optimal drug combinations. In fact one of the questions today was about leukemia’s. I just for, to shorten the talk I took out a screen with Retinoic acid which is used in the treatment of leukemia’s. That’s one example where we can apply this technology. Imagine now we have about a hundred and twenty drugs that are FDA approved. If we find a new drug and we want to say well what should we combine it with? Well, right now we only can do one against one twenty, very simple a hundred and twenty. But envision the future we’re going to have a lot more drugs. This technology’s going to be essential to identify new drug combinations. Here we have done it for Cisplatin challenging with seven thousand genes. This is what the results of such a screen looks like about three hundred and fifty hits with certain criteria. We integrate the data with predictive modeling coming from the Sanger and database. This was and I know there is a more recent release of this database, but here we can see familiar genes. For example, silence in BRCA1, one would predict sensitized cells to Cisplatin. We find many more targets possibly there could be combination treatment. In this case was ovarian cancer. We can also look at the TCGA angle and say well, we have hundreds of targets. Which ones are overexpressed in ovarian cancer that might be important to, for drug sensitivity? Here we can again limit through that type of analysis and hone in on the targets. The last few slides are going to be on matching patients with the right drug. This was from the New Yorker and really brings the point that Chris made. Many times DNA sequence doesn’t indicate the therapy. We have setup this idea to take fresh biopsies or pleural fluids, isolate tumors [indiscernible] challenges with all the available possible drugs. We generated beautiful drug curves using highthroughput equipment. Then we can match the patient with potential drugs. Right now our drug library that we have tested so far and these are not as updated data. But you can see that there are still a majority of drugs that are chemotherapies. But a broad number of targeted therapies as well that we’re testing. Reid Shaw was really a young graduate from Whitman College has been really pioneering the analysis of these high-throughput screens, showing the reproducibility of the technique of replicate. Also, here is one example; so far we have done about twenty cases. A lot of our cases now have originated from the Institute of Precision Medicine at Cornell. We’re still; we’re just actually validating some of those results in patients like xenograft. This is one example of a Gliobastoma that we had access to two different samples, one a biopsy one and two. The patient in between biopsy one and two received these new PI3 kinase inhibitor, BKM120 is also known. We could see that while the patient was sensitive in the first biopsy to the drug. When the patient relapsed under the drug it had acquired resistance. Interestingly it had required resistance to a number of drugs as well. In particular there was another drug that really had the biggest switch of resistance for this particular patient. Overall, this was again the first ten patients we analyzed with that library. It’s too early to say you know can we group cancer types? Can we group mutation? This is a, it’s a diverse set of samples. But I think it does show obviously that every patient is pretty unique. That there are certain drugs that do, are pretty effective. You see the blue drugs all across every patients. There we find highly toxic chemotherapies. They score high across everyone. Where are we going with this? The idea is to build in some way a large functional database that would address two needs. One is to pinpoint the drugs of the future and the drug targets of the future. These are the screen with the siRNA when we’re identifying hundreds of new genes as potential targets and matching those targets with known genomics and genetic features. The other part is really to being able to return drug matches to patients. Imagine a friendly database where patient and doctors could go and query you know with cancer type, with certain mutation and have a list of drugs coming out. This is where we would like to get to. This is the potential target that I think we’re seeing. I’ll take questions. I know there is a virtual audience to but certainly let me know if you have questions. >>: Can you backup a couple of slides? I’m just trying, so this is showing the change in resistance. Or is this just looking at, so you had the previous slide that showed the required… >> Carla Grandori: Resistance. >>: [indiscernible]? >> Carla Grandori: Yes. >>: Then now this is just sort of the same [indiscernible] across all the… >> Carla Grandori: Across more drugs… >>: These are [indiscernible] screens or drug screens? >> Carla Grandori: Yeah, drug screens. >>: Okay. >> Carla Grandori: Yeah [indiscernible] extension. >>: [indiscernible] distance of course a whole bunch of drugs. >> Carla Grandori: Yeah, so it acquires resistance definitely too multiple PI3 kinase inhibitors, to [indiscernible] kinase inhibitor as if you know the resistance is really… >>: Some generic thing. >> Carla Grandori: Yeah, broader, yeah. >>: Yeah, [indiscernible]… >> Carla Grandori: There is the only one genetic change which was amplification of an EGFR. But it could not explain this broad resistance to a lot of drugs. However, there were commonalities here that are not highlighted on certain sensitivity on the primary end that recurs that are unique to this patient and are preserved in the recurrence. In fact, we did identify potential drug options, so I’m showing only one side of… >>: Each of them are [indiscernible]. What happened? >> Carla Grandori: Yes. >>: But so what happened when you gave him EGFR inhibitors to the… >> Carla Grandori: There was an [indiscernible] change in the response to the [indiscernible]. >>: There was a what? >> Carla Grandori: Not much change in the response of… >>: [indiscernible] sensitive on both sides. >> Carla Grandori: On both, it gave some sensitivity, yeah. >>: I wonder if you could combine an EGFR inhibitor with the other drugs if that would have done something… >> Carla Grandori: If that would have worked. Yeah, it hadn’t, so the patient had also [indiscernible] amplification. There was a combination of the [indiscernible] inhibitor plus the PI3 kinase inhibitor. We did not test that combination you know before and after. But that would have been a good, so this is what we’re doing now for example. We have done about ten cases. We first test all the FDA approved drugs and the set of experimental drugs. Based on those results then we designed the combination tailored for the patient and tailored to the genomics of the patient, yep. >> Chris Kemp: Just to emphasize a point here that very few people are actually doing this in live, real time testing of patients cells first. Second, that the concept, what you mentioned was interesting and maybe you would just test each [indiscernible] receptor drug. But what we’re trying to say here is let’s test everything because we can do it and it’s not that difficult. >>: Yeah, right. >> Chris Kemp: Then it gives you a different perspective… >>: But you can’t test all the combinations because… >> Chris Kemp: No, the combination is too hard. >> Carla Grandori: No. >> Chris Kemp: But you can then get a different view about it looks like it’s more than just EGFR receptor and [indiscernible] broad resistance. What does that tell you about the biology of these tumors that we wouldn’t have learned before? That’s just; it’s a new way of thinking of testing everything. We’d like to call it [indiscernible]… >> Carla Grandori: Of course the important thing is numbers. We need to arrive to an important number of samples. But again we just started the new laboratory that’s doing these drug screens in patient derived cells has been open for two months. We have already about fifteen samples. We haven’t, you know we’re mostly working with Cancer Center and Institution. We have done very few patients because you know we feel that we need to have the genomics and not every patient has that information. >>: I mean one thing we have been working with Brian Druker [indiscernible] in leukemia as part of the treatment protocol. They can go there and sample the [indiscernible] functional screen against some targeted [indiscernible]. >>: [indiscernible] screens. >>: That’s right they do some [indiscernible]… >>: Not [indiscernible] as many. >>: [indiscernible] as comprehensive as what you’re doing. >>: We don’t have enough cells to do that… >> Carla Grandori: Yeah, I think if you’re… >>: How many cells do you need to do this? >> Carla Grandori: The challenge in leukemia is introducing the siRNA into the cells. For solid tumors they can grow as an [indiscernible] in layer and it’s easier. I think it’s more challenging. That’s why you know Brian’s lab has to use a complex technology that’s called electroporation to introduce the siRNA. For other tumors we just complex them with lipid particle and there are, they go into the cells very, very easily. But would there be robotic equipment that would adapt the electroporation in a large scale? There’d be no reason, now I think they do it up to seven hundred genes if I’m correct, so ten times scale up can be done. >>: They most take drug screens on patient themselves, right. >> Carla Grandori: Yeah. >>: They actually have data going back years doing that. >> Carla Grandori: Yeah, they are pioneers. They’ve been pioneer on that because you know Brian Druker’s story he was really the first one to experiment in vitro with this targeted therapy, and show how well you know they work and could predict in vivo response. >>: Carla, can you explain like if you’re developing drugs. You’re entering into trials. You’re going to test those [indiscernible] to drugs. How can your system accelerate that from drug trials so that we can get potentially drugs to market faster? >> Carla Grandori: We could by; you know knowing a priori which patient would benefit from that drug. You can envision making a very small trial that’s targeted and having high success from the beginning. That will increase basically the success rate of your phase one, phase two trials. By knowing whom to treat as opposed to go blindly across hundreds… >>: A hundred patients… >> Carla Grandori: Hundred instead of, yes, exactly. >>: Do you always need a trial? Or do you see it getting to the point where you just do it all kind of by doing the tests in the lab just to see what drugs effect the cell versus not? I mean, is there enough too really, or do you need the whole body for the tests? >> Carla Grandori: Right, I think for a new trial that has never been given for your patient that’s where the problem comes around. For drugs that have been given to a different tumor types it’s still not so streamline. But I see that there could be progress made there. For example, I had a patient from California just in December, just open. That was a friend of a friend and send cells, an ovarian cancer case with a BRAF mutation only usually seen in melanoma, rarely in ovarian cancer. The patient wasn’t treated with the melanoma therapy. That’s what the patient should have received. We did the test and it did respond to the melanoma targeted agents. That’s one example. Then they had to request the use of that drug but it was an FDA approved drug. The process can occur much harder when it’s an experimental agent, yeah. >>: In the wells that you do the testing you can put any kind of compound in those wells that you want. If you, if there were some sort of unique and untried compound from another country or from Europe that they’re finding success, but the US is just not even entertained. You could try that very inexpensively like why not? >> Carla Grandori: Yes. >>: To see if, oh my god, this black tea grown in the mountains in China has an effect. >> Carla Grandori: Yes... >>: Have you found anything that… >> Carla Grandori: Well, so there are… >>: Is naturopathic in nature? >> Carla Grandori: There are libraries that are called natural compounds. We had the screening facility that I setup. We purchased that library about a thousand. We screened several tumor types with all the FDA approved drugs, not for oncology. But also a thousand drugs and a thousand natural compounds. Give you an example on a lymphoma, set on lymphoma was found that in the derivative of avocado oil was highly effective. Now the difficulty on moving such a finding into the clinic it’s challenging. Because we don’t know what the target is. How you’re going to convince a drug company to do a trial. Then often these natural compounds are very complex so it’s not one molecule. That’s difficult but you know if one was I think one should try. >>: [indiscernible] one question. >> Carla Grandori: Yeah. >>: What’s the correlation between the in vitro and in vivo response? You know so if you get something that’s going to grow… >> Carla Grandori: Yes. >>: In these in vitro tests. How well does it actually translate [indiscernible]. Do you have any data on that? >> Carla Grandori: Yeah, so we don’t have any data yet. What we can say that for example from our huge siRNA screens. That we honed in on the Wee1 and it worked in the mouse, and in the clinical trial, in the first patients enrolled there are good news. That now we cannot share but. That has gone from the screen to a human. For CK1 Epsilon we haven’t gone to the human. We have gone to mice. We have, so far the one that we tried are top candidates. Is not and usually they’re top candidates as a result of a screen together with a lot of validation around it, in vitro. They usually translate in vivo. I think if we were to take all the raw hits I don’t know what percentage would. But, so I think definitely the validation is necessary from many angles before. >>: Even for patient do you need to try, do you envision having to try all the drugs for that patient? Or will that step go away I mean when you have enough data it will work? >> Carla Grandori: Yes, that step will go away. That’s where we… >>: Because is it inherently expensive to try all these different things? >> Carla Grandori: Yeah, it is and it isn’t. I think the test that we give today is not going to be the test of tomorrow. I haven’t brought my trays. But right now we test four hundred drugs in one tray, different concentrations for each. I envision again that, pretty soon we’ll know based on the genomics of that patient likelihood that these are going to be the best combinations. We’re going to test all the combination for that patient. We are looking to purchase this dream machine that can really do the combination for us and is driven by sound waves. That’s what we had at the ThinkTank. >>: Yeah. >> Carla Grandori: It costs about three hundred and seventy thousand dollars. But you can easily program to do that. >>: [indiscernible] lab site. >> Carla Grandori: Lab site, yeah. >>: From [indiscernible] patients. >>: Yeah, it’s for a thousand drugs there’s a billion combinations… >> Chris Kemp: Well, no, you would nail it down… >> Carla Grandori: But you’d tailor it to the genetics of the patient. >>: That would be very, so we’re going to learn the… >> Chris Kemp: You’re going to have to… >> Carla Grandori: You have to… >> Chris Kemp: We’ve only done less than a hundred of these experiments. Once we get up to several hundred we’re going to start learning the rules. Then we can throw out ninety percent of the drugs… >>: That [indiscernible] >> Chris Kemp: Yeah and then focus on a usable number. We want to learn the rules. Our vision is to get that map that I showed and sketch it out completely and learn the rules. How to navigate for this cancer with that mutation this is a better drug than what the patient’s currently having. That’s our vision. >> Carla Grandori: Yeah. >> Chris Kemp: If we make a step in that direction we will be enormously successful. Because right now I showed you the, you know the pancreatic cancer situation is one example. That’s the vision. >> Carla Grandori: But we, so I think we’re, I am thinking in the near future for example. We’d like to focus on one cancer type. We’re DNA sequencing information for each sample and do a hundred or two hundred. Then test all the combinations because there’s a new drug entering for breast cancer now. It’s an FGFR2, FGFR inhibitor, Fibroblast Growth Factor Receptor. Eli Lilly, Pfizer, several, AstraZeneca, have inhibitors against these drugs. It looks like about ten percent of breast cancer, even triple negative for which there is no target has that aberration. That co-exists with PIK3CA mutation, P53, and often MYC amplification. That’s the picture we’re looking at. We are taking those types of samples and testing with all the combinations including all those new FGFR. I’m sure we would find an optimal. I think in the future the tests will be much simpler. Maybe it will go away altogether, but. >>: How is what Microsoft is doing in your research dovetail with what Carla and Chris talked about today? Is there an overlap or are you following toward the same goal? >>: There’s a number of things we’re doing. You know I, so the stuff of the OHSU folks of the [indiscernible] Project there’s definitely some similarities, right. They’re trying sets of drugs on patient [indiscernible] cells and looking at the genomics of what’s going on in the tumor, to try to understand relationships between drug efficacy and the tumor. They’re trying, in particular they’re trying to come up with combinations since we spent some time helping them select candidate combinations to run in their screens which they’re doing now. >>: You’re looking at genomic data? >>: We looked at genomic data. We also looked at basically just functional data by you know the IC50 kind of hit for how effective the drugs were at killing cells and some RNA [indiscernible] too. There’s that. The rest of the stuff we’re doing isn’t directly applicable although… >>: I mean does Microsoft feel like there’s a software play in this realm of research? Or a, I mean a computational science for sure but where’s the kind of the Microsoft I don’t know business plan, or… >>: Business plan, well we’re in the research lab so we’re not response… >>: You don’t have to… >>: You’re on the business side it’s more around you know what’s the big influence you know just [indiscernible] machine learning in a very general way [indiscernible] broad. There’s definitely some [indiscernible] in that direction which would tie in to this. But again it’s really [indiscernible] infrastructural. >> Carla Grandori: Yeah, I think is, and you know we talked about combination. But it’s doing these broad dried sensitivity tests and then see if there, we’re starting to see some pattern. But you know we only, again, now have the twenty. But we’re, there are certain two more type exquisite sensitivities to certain drugs and other don’t. If you know I think that would be the type of data that one could feed to the machine learning type of approach. >>: Does the siRNA, so there’s kind of the mutational state, the genomic state. But there’s also the epigenomics state. Like between these two would it definitely be a mutation or could it just be a difference in the epigenomics state? The siRNA screen kind of can capture both in some sense then… >> Carla Grandori: Yeah, yeah I think you absolutely hit it right, yeah. The siRNA can, is almost, is readout of the epigenetic state as well of the cell. Yeah and so it would be, again, it would be nice to know and particularly for these patient how that would integrate. >> Chris Kemp: That’s a great question also, so we also have a library of epigenetic genes that regulate the epigenome. There’s about a few hundred of those. It turns out that in pancreatic cancer for example there are a number of mutations in these genes that bring like the epigenetic landscape of pancreatic, of the chromatin structure. The [indiscernible] H two is an example of, it’s a potential drug target and it’s coming out of our screens. As well as, the BRD4 gene that we talked about in pancreas, in the men cancers also regulates chromatin. I think this is an exciting new frontier. The question of genetics could play into this, too, because how do you have this global resistance to all these different drugs which have different mechanisms in action. It’s probably epigenetic… >> Carla Grandori: Epigenetic, yeah. >> Chris Kemp: Adaptation of these cells to stress in general. It, so this is a great frontier that I’m really excited about. Yes, the RNAs screens they have no preconception with what you’re looking at. That’s what’s cool about it. It just tells you what does the cell depend on to live? If it’s epigenetic or genetic it almost doesn’t really matter in some cases. >> Carla Grandori: Yeah, it would be nice to have a better protein atlas I think to have because there’s sometimes signaling molecules might not be changed dramatically in expression in their activities different. That’s another level also that’s hard. We have looked a little bit at the RPPA database. But just found a, this for different reasons because we’re part of the MYC PAN Cancer Analysis Group. I didn’t find it as informative or accurate. That would be, but the mRNA certainly is something to be able, we should integrate not just DNA sequence, yeah. >>: Might that be genetic genes typically hard to drug because the [indiscernible] are active in the nucleolus? It’s hard to get drugs in the nucleolus, right, like TFs are hard to drug too. What do you think? >> Carla Grandori: Yeah, I think more the TF are hard to drug because they don’t have, you have a small anchoring pocket like BRD4 has, or an enzymatic activity that can be inhibited by a small molecule. MYC for example forms a heterodimer in the cells with MAX. It’s a large surface that interacts. A small molecule will not budge that large, the large surface of interaction, will not disrupt it, so people have tried quite a bit. There is really inherent structure of transcription factor and the lack of docking sites for small molecules, but, yeah. >>: Just [indiscernible] is a good drug target. >> Carla Grandori: Yeah. >>: BRD4 is a good, is potentially a drug target… >> Carla Grandori: Yeah, so these are the exceptions, yeah. >>: [indiscernible] .1L company called Epizyme that’s making drugs too is epigenetic. There are some you can… >>: Whether there the cancer ones, but [indiscernible]. >>: Some you can, some you can’t. >>: Right, there’s a couple of drugs out there for ID20 and H2. >>: Yeah. >>: But those aren’t really that exactly that epigenetic brand. >>: No. >>: The metabolic genes of the phyt… >>: Correct. >> Carla Grandori: Yeah. >>: They’re phytoplasma genes. >>: Right. >> Ravi Pandya: Do you want to thank? >> Carla Grandori: Yeah thanks. [applause]