 Okay, we'll have people come in soon, but it's my pleasure to welcome Dr. Leppert. John Leppert is an assistant professor of urology. He, his clinical practice is at the VA, but he's a very integral part of our kidney cancer program here at Stanford. John trained at UCLA and then came to, he is also trained in minimally invasive disease and does most of his surgeries laparoscopically, just like Ben Chung that you earlier heard from. John is going to, his research has been in looking at big data basis and data sets and today he's going to talk to us about using big data to improve kidney cancer. John has done a remarkable job of bringing together good minds here at Stanford who are interested in kidney cancer and it's almost like a think tank that comes together once a month. Two times a month. Two times a month so that people can come together with who are engaged in the same cause. So it's a pleasure to have John talk to us on using big data to improve kidney cancer care. Thanks John. Morning everybody and thank you Sandy for the introduction it's great to see so many of you here and I hope in this little talk to give you guys the 30,000 foot view over big data and some of the research efforts that are gonna be I think in the near future impacting the care of kidney cancer. So like Sandy said I'm a urologist here at Stanford and at the Palo Alto VA and my research kind of lives in two worlds so I wear two hats. One hat is looking at biomarkers so how can we better diagnose cancer and identify treatments for cancer and then the other hat I wear is epidemiology which is looking at these larger data sets so that's what I was gonna hopefully share with you today. In terms of disclosures I do have grants but I don't have any industry funding so a lot of my grants are related to the topics of what we'd be talking about today. And for those of you who have been to prior kidney cancer support groups and have heard me talk I always include this slide so this is my grandmother at my wedding and this is one of the reasons I became so interested in kidney cancer so during my training as a urologist my grandmother had a fall and the fall led to an ultrasound which led to an MRI which led to a CAT scan where they discovered that she had bilateral kidney tumors that were actually quite large and her fall was in her late 70s and she's now almost 10 years cancer free after two surgeries, one for each kidney. And if I can have just a sliver of her tenacity in terms of over the course of my career I think it will be really beneficial. So starting with just the beginning what is big data? So you'll hear more and more about big data I've even seen commercials now for IBM Watson where they show a doctor's interacting with the computer and they're talking about big data but what I love, I love this quote because big data can be almost whatever you want it to be and this is a famous quote if you see any talks about big data it talks about how big data is like teenage sex everybody talks about it nobody really knows how to do it everybody thinks everyone else is doing it so everybody says that they're doing it, right? And so that's really what big data is it's this generic term that encompasses anything that's big and big data existed and is probably far more sophisticated in a lot of fields before it came to medicine so if you imagine going on Amazon and buying something there's an algorithm then that recommends additional products for you based on what you searched for based on what you've bought they know where you've been they know what you've clicked on and they use that to sell you stuff and that is basically big data think about the millions of people that use Amazon and how they then parse that into a product recommendation and that's true for many different things social media, Google I'll give you an example just from me so I live here in Stanford Google is my internet and my cable TV provider Google's my default search engine and my family uses Google Express to shop for stuff so if we wanted to buy something and have it delivered to our house Google knows what we've bought and that data becomes incredibly valuable to Google so that they can target ads to me and sell me other things and I've been willing to trade that information to them for the convenience of the services that they provide but this is the world that we're gonna start living in where all of this data is integrated so what could big data mean then in medicine? Well the earliest I think examples of big data in medicine would be looking at a large number of patients so if you took the National Cancer Registry Seer you could look at 100,000 or 50,000 kidney cancer patients and as a group see how well they did and that's really just the beginning because that's one dimension if you think of a graph where on one end of the graph there's a tick for each patient and we're looking at 10,000 patients you can imagine then that the Y axis could be something about that patient so now let's start talking about bigger numbers what if we sequenced the DNA from each of those 10,000 patients and so that's about three billion base pairs of data one or zero for three billion spots so now 10,000 people on the X axis a billion points on the Y axis and now let's talk about making it even more dimensional which is well what if we do it more than once? What if we do the sequencing when they're diagnosed and then when they start treatment and after they've been on a treatment to see what changes or what if they fail a treatment? So if you look at what has made this big data then possible the obvious example is genomics so in 1990 we launched the human genome project so that was an attempt to sequence the DNA of a human single human it turned out to cost three billion dollars to sequence the DNA from one person and so that involved 20 universities, multiple companies more than five countries a huge effort to sequence one patient and now if you look at what the cost has been to sequence a patient over time you can see that the cost per base pair so that one or zero and the cost per sequencing a whole genome for the whole patient has plummeted it's about a thousand bucks now so we've gone from billions of dollars for one patient to the ability now which is actually outpaced Moore's law so for those of you who are familiar with like storage and your computers and how everything gets smaller and faster over time this is actually accelerated quicker than that law would have predicted so it's becoming cheaper and now feasible to do this and there's been companies then that have done this explicitly so you'll hear ads for 23andMe or Foundation One or all these companies which are offering services in addition to what you can get through kind of traditional medical establishments to sequence your DNA and so as we start looking at this highly dimensional data where we have lots of data points, lots of patients we need to then have the tools for how we can decipher that data and actually use it for a benefit so I'm gonna give another example being here at Stanford it's probably similar at other large institutions my daughter goes to public school she went to kindergarten and she's in a kindergarten class with a friend her friend's dad happens to be the chair of genetics here at Stanford so there was a virus that went through our kindergarten class and our daughter was homesick with a high fever for about a week and it turns out about half of the class was too the virus spreads to the rest of the kids in our family to me, my wife, everybody's sick we're miserable and my daughter's friend's birthday party gets postponed because everybody's sick so when they reschedule the birthday party I go to the birthday party I talk to Mike Snyder my daughter's friend's dad and I ask, that was an incredible virus I mean I felt horrible I wonder what it was and so he said, I can tell you what it was he's like, well what do you mean? he's like, well I sequenced it so it turns out that Mike Snyder for two years had been doing multiple tests on himself so he had been doing full genomic sequencing proteomics, metabolomics all of these omics on himself over the course of time and when he got sick it happened to be during this study so he did everything again and what was really interesting in this case is this was a respiratory sensitial virus and ours a common cold kind of virus that everybody had gotten and gotten sick after that virus for him he actually developed an antibody against the insulin receptor and became diabetic he had no family history of diabetes he's not obese, he's active he rides his bike past my house on his way to work every day and so this was a really interesting kind of experiment where he had tested himself at multiple time points and just a funny coincidence that we happened to include the virus that my daughter got in kindergarten which gets published in Cell because he had sequenced it and so as we start to do these omics approaches genomics and proteomics and we start to get this data that we've never had before I think we're gonna learn a lot about what actually happens and what changes over time for our patients another thing that has really facilitated this big data approach oh no, fire drill should we check outside? all right, so I think we're probably then okay so another thing that's facilitated kind of big data approaches is that we have digitized the medical record so in the old days you'd have a paper chart sitting in your doctor's office that chart would have handwritten notes about what you had told the doctor what the doctor had recommended the handwritten prescription perhaps to the pharmacy would be in there and these are all being converted to ones and zeros so I'll bring you back to where I practice clinical medicine which is at the VA so the VA is actually the largest national integrated healthcare system in the United States and it includes more than 150 hospitals it covers nine to 10 million people per year depending on the year and that has a single electronic health record and that health record captures everything that happens at the VA for each of those patients so for those nine million patients we know since the mid 90s every lab test and every lab test result they've ever had every prescription they've ever picked up whether they've refilled it how long they've been on it every surgery they've ever had every diagnosis that a doctor has given them gets coded into this electronic data and it's to the point where for just the lab results the VA records about one million lab results a day that get added to this record and these all sit in a big data storage warehouse in Salt Lake City and Austin, Texas and these places where then it's accessible to researchers so if you wanted to do a study about people in the VA you can parse through that digital information where in the past that paper record made it so that everything was isolated to the chart that was sitting in your doctor's office and the only way to get at it was that somebody would have to look at it there's a great example then that of what this could mean in the future and this happened actually at the children's hospital here at Stanford 13 year old girl came in with lupus and she had multiple risk factors for blood clots but the literature and like the textbooks don't say exactly whether or not that patient should be on a blood thinner and blood thinners carry risks so you could imagine a button that the doctors would have in the electronic medical record that would say how do patients like this patient do or what are the outcomes for patients like my patient and they envision this kind of green button approach don't know why they've called it a green button but the idea that the electronic medical record the doctor looking at the computer can get the data of what happens to similar patients sorry guys my talk is on fire did I commit a felony did I say fire when the fire alarm was going off I'm gonna peek outside just make sure all right so we've cooled off and so this report was actually an example of where this pediatrician clicked on a button to do a search of data that was then in the medical record at Stanford where they could look at they found 100 patients like this one and they decided to give blood thinners based on the outcomes of those other patients so you can imagine that with the digitization of an electronic medical record we may get closer to what you envision with an Amazon type search and a product recommendation or a Google recommending another site for you another example of how things have been digitalized and that's imaging so it used to be that they would develop your X-ray it would be on a piece of basically a photograph on film that would sit in the jacket in a big warehouse and they'd look at it and then it would go back onto the shelf well now all of our imaging is almost all digital and these are ones and zeros that can then be parsed by a computer and you can imagine that a computer could try to do a better job than even the radiologists or identify features that the radiologists might miss and do a better job at diagnosing a mass in the kidney or an area where kidney cancer could have spread outside of the kidney so that kind of brings us to the next set of definitions so we've heard about big data we've heard about omics and how are we gonna then implement these approaches and so there's two kinds of catchphrases that people have used so one is personalized medicine and one is precision medicine for my mind they're pretty interchangeable but there's some subtle differences at least in terms of how people have interpreted them if you could imagine for personalized medicine that I'm going to find out exactly what kind of tumor you have maybe know more about that tumor in terms of its genetic makeup find out that it has a certain gene mutation and then give you a drug specifically targeted at that gene mutation and to try to give you a better chance of response because I've now personalized your therapy so instead of treating everybody with kidney cancer the same we're going to start being very patient centered and each patient is unique precision medicine is similar but the idea then is a little broader which is we wanna find the information that helps us categorize patients into groups that will then respond so we do that a little bit you have clear cell kidney cancer you have papillary kidney cancer you have a different type and we can create buckets where maybe we know how groups of patients will do and we can better define those buckets and better characterize with more information and the more information we have hopefully the better we are at predicting outcomes and how people will do so there's been a lot of really exciting things that have come out in just the past couple years and a lot of these are directed towards cancer and kidney cancer is going to be part of this for sure so the precision medicine initiative is one of these so this was introduced in the state of the union address in 2015 and for this fiscal year the government put a little more than $200 million towards this effort and this money is basically going to create a cohort of one million patients and in these million patients you'll have medical records you'll have the ability to do genomics and other kinds of testing that we would consider to create this big data resource of these million patients and one of the first goals of this precision medicine initiative is to figure out how to do better for cancer patients but this is also more than that and that is you'll hear a lot of criticism for how we do research which is looking specifically at disease states this will also look at people that are healthy so what is normal? What is a normal variant? Maybe you have a gene mutation but it's really not pathologic doesn't cause disease, doesn't cause you any problems and so we need more information to try to figure out what the bad actors are as we start getting this information so this is accruing now multiple sites across the whole country and trying to create this cohort if you're interested if you go to the whitehouse.gov or to the National Cancer Institute website there's links where you can try to sign up and they'll do a survey and they'll come to your house and do additional data information collection and they'll do the samples they need to do if you wanted to participate so that was the 2015 State of the Union Address the 2016 State of the Union Address introduced another thing which is this cancer moonshot idea and so this is really trying to accelerate the cure for cancer and this is a group effort across the whole country Biden is kind of chairing it since his family has been touched by cancer recently and you see as you look at these efforts all of these cancer centers trying to come together for this common goal and we were actually talking about this before the talk one of the goals of the cancer moonshot is more than just finding a cure it's also trying to get access to the right medicines for patients no matter where they are access to clinical trials and freeing this data so one of the limitations for using this big data is that the ownership of this data is in question if you have your tumor sequenced by 23andMe or another company or Stanford is that information yours because it's your tumor or is that information whatever the property of whoever obtained the information and so how can we break this data out of the silos that it's at the same silos that were capturing data in the paper chart in a doctor's office how can we then share all of this information so that we are better at using the information specifically like we were talking about for rarer tumors where maybe there's only a few tumors at one institution that are of that type and if we leave them there we're never gonna get enough information to really act on it and do a good job so part of freeing the data is one of the goals and priorities of the cancer moonshot which is just announced this is still in the very much the early planning phases but I hope that in the near future these are things that become a reality and back to the VA so the VA actually has created to date the largest genomic resource in the world and they've enrolled more than 500,000 veterans to be part of the Million Veteran Program so their goal is one million and these are large collections then of these resources and one of the beautiful things about this data is it's linked to people's use of the VA healthcare system so they have all of this additional data that can then supplement the sequencing of their genome or of their tumor or whatever else then is part of the Million Veteran Program so I wanted to give a couple examples then specifically of big data in kidney cancer just to kind of wrap up so the Cancer Genome Atlas is a consortium group that has been sequencing different types of cancer and this was a paper that came out in 2013 where they had sequenced a few hundred I think it was about 300 clear cell kidney cancers and in the future 300 tumors may not be exciting but when this came out this is a big deal and it's the largest resource that's been published to date and I want to just show these plots data visualization becomes hard when you're talking about 3000 genes that code proteins and three billion base pairs and 300 patients but one of this plot right here it just lists the gene mutations where blue is whether or not they had a problem with that gene for each patient and so you can see that top line has a long width of blue and those are VHL mutations so loss of VHL which is a gene you guys may have heard about which codes for a protein is one of the most common events in kidney cancer and that really defines clear cell kidney cancer so that we know that about 60 to 70% of patients that have clear cell kidney cancer have this mutation and that's one of the things that triggers all of the downstream effects for kidney cancer of this type but you can also see that there's seven or eight other mutations there that people don't think of as dominant but that occur with enough frequency that if we knew about it maybe we would use drugs differently or maybe this, let's say it's a BAP1 maybe BAP1 is the Achilles heel of that tumor and if we were able to target it specifically with the drug we'd do even better in terms of how we would treat that patient more data visualizations from that same effort where they look at how genes proteins change over the stage of tumor over the grade of tumor and it's more than just the gene and whether or not you have an A instead of a T at that one spot genes can get turned on and turned off by epigenetics so things can get added to the outside of genes that make it so they don't get transcribed and so the more we go down the rabbit hole the more we learn how complex this problem really is and the way that we used to think about biology is becoming more and more complicated the more that we learn and so back to the VA as an example I'll just show you a little piece of work that we've done in our group we have 150 hospitals worth of patients and we can search those records to find all the kidney cancer patients and then we can take all those kidney cancer patients and find everybody that had had a kidney cancer surgery and then we can look at everybody who had had a kidney cancer surgery and look at their kidney function on their lab tests over time so this was an example where we had about 14,000 kidney cancer patients that had had a surgery and we were able to link that with almost half a million creatinine measurements so these are blood tests to measure the kidney function to know what the kidney function did over time and the question is if we take your kidney out or if we take part of your kidney out what can we expect that your kidney function will do over time? Are you gonna be at risk for dialysis? Are you gonna be at risk for other side effects of having less kidney function? And the way that we plot this this is a contour plot of those 400,000 blood tests we actually see the blue line is people that had a partial nephrectomy and their kidney function is higher than people with a red line they had a whole kidney removed, a radical nephrectomy but one of the very reassuring things that we found from this study is that the kidney function was very stable over time so this is 10 years at the far right of that screen and that the average kidney function over that time actually stayed very stable so kidney loss of function after surgery behaves differently than people that have lost kidney function for diabetes or for high blood pressure or from some of the other things that then progress on to end stage renal disease and transplants or dialysis where in these patients the kidney function actually was very stable after they had had their surgery they found their new baseline and that's where they stayed in this case 10 years What's EGFR? Sorry, estimated glomerular filtration rate basically how much is the kidney filtering and more is better so in this case if you have a higher GFR your kidneys are working better than if you have a lower GFR there's lots of jargon, I'm sorry similarly imaging, I didn't have time to add it but there was a paper actually that came out this week and this is a paper from last year where they took sequencing and then they digitized the images from a CAT scan and then they asked the computer to find differences in the computerized CAT scan images that would predict how the patients did and with that computer information so you might hear about machine learning artificial intelligence, deep learning these other terms for how they parse computerized information the computer was able to detect features on the CAT scans that were better than what the radiologists had said on their own at predicting outcomes and the paper that just came out this last week they did the same thing with digital pictures of people's pathology slides so this was for lung cancer and so they were able to look at lung cancer pathology slides the pathologist gave it a score and predicted whether or not this cancer had a high chance of recurring and then the computer did the same thing and the computer was better than just the pathologist alone and could pick out features that you don't appreciate to the naked eye and as we digitize a lot of these big data resources I foresee this being part of the future where we're gonna have computerized algorithms to predict better what your type of cancer is on your CAT scan or on the pathology slides after it comes out what is the chance that this comes back? Oh, so these were genetic terms looking at the expression of RNA so I don't want you to get hung up on it because it's not important but the idea was they had the genetic information and on top of that they were able to computerize the CT scan so the more aspects of the data that we have the better that we're gonna do and here's kind of my favorite I get emails because you may get these too that basically are the lay news reports and these come out all the time aspirin's good for you, aspirin's bad for you a glass of wine a day is good for you, it's bad for you there's a comedian that asks a joke where he starts his routine and just says milk, is it good for you? and he just pauses because the audience doesn't know what to say it used to be when we, I'm old enough to say when I grew up of course everybody had a glass of milk that was good for you but now we worry about 2%, 1% skin milk all of these other things a lot of the data that's out there especially the data that makes it to the news and these news reports are sensationalized and poorly done right and so coffee, good for you, bad for you who knows but this was within a span of just a little bit of time in my own email feed of these news reports this particular drug causes bladder cancer this particular drug doesn't cause bladder cancer and so the truth is a lot of these studies are limited and they're limited because we don't have access to this big data resource they're often done at one center they're done with a small number of patients they're done with a very specific group of patients things that aren't generalizable so here's a way to round it and that's big data so this actually just came out this past week as well where they took everybody in Sweden Sweden has a birth to death national health care system so it's a lot like the VA except bigger and that they capture every drug you've been on every lab test you've had every surgery, everything you've ever had that's part of your medical record from your birth to your death as long as you're in Sweden and they looked at every single drug that was in that resource and said does this drug have an association with cancer? Any cancer is one analysis prostate cancer, breast cancer, colon cancer they started going down the common cancers they didn't do kidney but this is the type of data that is needed to really say something with some authority in a retrospective way about whether or not coffee causes cancer or a drug causes cancer we need this kind of data we need to do it in a very methodologically rigorous way and a lot of what you see in the news just doesn't meet that criteria yeah thank you right so Bill if you don't mind I'll paraphrase your question which is a lot of bias a lot of influence I started my talk with the disclosure slide a lot of things that influence you will then influence your perception of the data and so absolutely right and as we do research with big data it is so big that you will find things that are statistically significant it's just a matter of if you look and some of those things won't make any sense you'll find data that if you eat pomegranates you have an outie belly button and that will be statistically significant in this data set because you're looking at 10 million people but it means nothing there's no clinical reason to say that that link occurs so you not only have to be smart about what questions you ask and you have to do the methods in a rigorous way it has to then pass the sniff test it has to have face value that this is possible and then once you have that piece of information you can't stop there you have to validate it you then have to test it prospectively and so for kidney cancer in a very kind of to try to bring this back specifically if we found a gene that was associated with kidney cancer outcomes looking at patients that had been in one clinical trial then the next step is we have to recreate that study with a different set of patients and we probably are gonna try to work on a prospective clinical trial for that gene mutation and that new drug that we're giving to that patient group so that we can balance out the bias and we can try to do the best that we can and a lot of what you're gonna see in the news was way before those steps have ever occurred so I'm gonna end my talk then with a slide so this is a slide from the Cassini space probe where it passed Saturn and it's looking back and that little dot down there is Earth and I think that that's what we're gonna do when we look back at how we treat kidney cancer maybe 10 years from now, 20, 30 years from now just like we look back at leeches and the middle ages of surgery and our goal really is to stop treating everybody the same to personalize or have a more precise way to deliver drugs, to deliver surgery, to deliver the right care to the right patient at the right time and the better that we can do with categorizing patients and getting information these big data type resources I think the better that we're gonna do for our patients. So looking forward, we need what turns out to be a much more diverse group of researchers we need people that are familiar with computational science, data science we need the biologists and the people that are in the lab doing the work we also need the clinicians and we need the patients that we need this data then to be freed and freely flowing between the basic research groups at the different places and we need to then implement that data in a rigorous way and I think that's what's exciting and what's down the road for kidney cancer in particular in the very near future. So thank you very much for your attention and I'm sorry about the fire drills. Thank you for the information. I have a fairly simple question but as the bins for precision medicine get filled does the filtering become more granulary to make it more personalized? Let me make sure I'm understanding are you saying that when people enroll in those cohorts and the cohorts fill up? Well you mentioned that right now we use more of a precision technique where all the patients that had clear cell go in one bin. Well over time as that bin gets filled and you start to see how different patients react to different medications or whatever, does the filtering become more granulary? So as you get more information on these different patients you can apply it to new patients in a personal method. Yes and so I'm gonna take that to the extreme which is there's this concept of a clinical trial of one patient, right? So as we learn from prior clinical trials and we can take that bin and let's say the bin starts at clear cell that bin will then get divided into clear cell type one type two, type three, type four based on those mutations the bins are gonna start to get smaller and smaller and as we have more and more experience we'll know how people have done with the 10 drugs now that we have for advanced kidney cancer in each of those bins to the point where we're eventually going to have a single patient that comes in they have their tumor sequence let's say or some other kind of information that helps us decide what bin they fit best in and then we give them the drug that we think is best and we can look at whether or not that works and so as we become more precise in how we categorize patients I think you're right, we're gonna become more personalized and the ultimate will be that each patient is treated individually based on those categorizations that we can do and that clinical trial of one patient is kind of that idea. Hi, so two questions hopefully useful for everyone. So for those of us who are very technologically inclined I saw your chart of the genome sequence dropped by a factor of one million or so but as you know it takes a decade maybe two for that to have any tangible improvement for the average patient so how does someone avail themselves of these new technologies or the trials or whatever could be done earlier and especially in the case of the genome if it's only a thousand or $2,000 there's some way we can make that useful for ourselves in the short term. So one is you can always be proactive and these companies now 23andMe Foundation One make it so that you can do that on your own and without the need of your doctor you basically can request your pathology slides they get sent to a company like Foundation One and they will do an analysis. At this point I think a lot of times that might be a waste of money if we don't have a way to act on that information so you may be able to get the information it may not be actionable yet but you can definitely do it and if you wanted to be proactive you could do it now on your own. The second thing I wanted to say is I don't wanna trivialize the whole genome sequencing because you're right the cost makes it easy. What's actually now the trick is the data analysis of that sequencing and then acting on the data. So yes you can spend $1,000 and you can sequence it and then it's gonna sit on a hard drive and it may take some postdoc or some data scientist a few months then to parse through that data to find out what your sequencing actually looked like and that's actually I think gonna be the rate limiting step for a lot of this work is once we get the actual data how do we then interpret it and act on it? In the future what that may mean is as we develop new drugs they're so slow to come to the market right? If we develop a drug for lung cancer we may be able to repurpose it to kidney cancer to bring it to the market much quicker especially for patients that have failed traditional therapies and so I hope that in the near future maybe even shorter than 10 years from now that the process to approving drugs especially for specific indications becomes much quicker because we can repurpose things that are already in the pipeline as we learn more about mutation and you'll actually see this now and you might hear about breast cancer and Angelina Jolie having a mastectomy because she's BRCA1 positive right? And now there's whole groups of researchers who are looking at BRCA cancers which may be breast, prostate and these other cancers and they're starting to try to treat them in the same bucket no matter where they came from and that may accelerate some of this drug discovery. But you're right, there's still a moat between you and then getting the data and then acting on the data might even be another moat another step that we haven't really figured out how to do it and scale. So and along those lines you had the example of your daughter's friends, father runs the general department and he can sequence anything he wants to. Are there open source tools or the equivalent available for an average person to just start to tinker around with that? That's a great question. With the sequencing you're still technology driven because you need the technology to actually do the sequencing and the computational power to do that. So maybe a few people might have access to that but that would be really rare but the other things may be quicker. So let's say your electronic medical record. There are companies, Google Health tried this but I think they've folded that would then store your medical electronic health record and in that data then I think that is much more the barrier in terms of doing your own analysis of your own health record or groups of health records if we've compiled things into a resource. I think that is actually within reach and that's something you could do with kind of off the shelf kind of stuff and more of the DIY kind of analysis. Right, so once you had the sequencing, absolutely right. In fact, a lot of the computational work is R-based, Python-based and that's actually usually out there in public Githubs and so once you had the data you could actually pull out the resources and if you had the computational power look at your own sequencing, that's true but you got to get the data first and that's the trick, yeah. So the bottleneck is getting the data? Getting the data and then getting it to you so that you could look at it yourself, yeah. Thanks. Okay, thank you so much, John. It's amazing. Thanks for all that you do for our community and for questions.