the one thing that's always driven me is to try to find out what is the question that really needs to be addressed between the cellular lab and the genomics lab digital pathology all those things what do you see as a potential for uh what we're building well the the potential is that you can start to assemble pieces based on what the question is that needs to be answered and you can approach it in a novel way not necessarily replicating what someone else has part of the challenge with Co I think is people were actually seeing the scientific process as it occurred which is why the information kept changing where do you think we are right now with AI not just in health but as a society AI can be very useful but it also can be very much abused that's the concern I have technology is the instantiation of a solution to yesterday's problem and doesn't know what today's problem is Michael leedman welcome to the podcast thank you really excited to to have you here today and for the for the listeners go over Michael's bio Michael's I consider a good friend and and Mentor uh he is one of the best thinkers I've ever met and you've taught me how to think a lot over the last six years it's been eye openening and I I try to take the lessons and you know every time I talk to you I learn something about a way to think or uh uh how to Grapple a question and um it's it's it's been eye openening and we we've been talking about critical thinking this week for instance and and the importance of it but um you know for the listeners uh Michael is uh on our scientific Advisory Board um he was the chief scientific officer uh for United Cancer Centers uh when we were launching that unfortunately covid happened and kind of stop that uh stop that launch but um he has a PHD in theoretical chemistry and protein crystallography I still don't know what protein crystallography is um uh he was let's go over the bio here um he was the director or he is the managing director of ipq analytics and strategic medicine um he was the executive director of the windbur research institute um he was uh he's an Adjunct professor of pharmacology and Physiology at Drex um he was the uh director of computational biology and bioinformatics for the University of Pennsylvania Cancer Center uh he served as the global head of computational genomics for RO Pharmaceuticals and director of bioinformatics and pharmac genomics for wyth pharmaceuticals uh Michael welcome to the podcast thank you I could have kept talking I could have just kept going on the the bio because it's even even longer but um yeah thanks for uh coming to Nashville well I appreciate the opportunity and it's always good to see you yeah absolutely um so uh for the people listening um want you give a brief overview of uh your background how how you got into science and kind of how your career unfolded uh over over the last I guess 50 years well yeah it's been that long um so my background is a bit unique because I going back and forth between industry and Academia several times uh I started out with an academic career like most people uh at uh Institute for cancer research and then went on to Mount Si as a faculty member um but what I've always been interested in uh is the issue of disease and helping understand disease to improve patient outcomes and um while I have formal training in chemistry I've never been in a chemistry Department um but I've been able to take advantage of a lot of different uh opportunities to learn pieces from them learn different perspectives from them and try to assemble that so that together to address problems that I think uh are complex problems they're hard problems but I think we really need to start to look at hard problems because they're not going to get easier on their own and uh some of the issues I see today are that we're tending to try to use technology to address problems without always understanding what the problems are and um the one thing that's always driven me is to try to find out what is the question that really needs to be addressed um so even when I moved into industry um what one of my jobs was to go back to academics and and find new technologies and so on and I explained uh uniquely I went to Amo the oil company when they were in biotechnology that showing up on an academic campus coming from industry people would want to know how many bags of money I was in and that wasn't the goal the goal was to really look at the science and form relationships so what I uniquely carved out that I recommend to a lot of people is that I could also have academic appointments when I was in industry and that way we collaborated we looked like scientists which we were but we also could publish periodically and that enabled me to go back and forth much more readily than if I had only going into industry and not have that um set of academic credentials so to speak yeah so for for the people listening you know there's uh there's a few different parts of of science but you know you have industry and then you have the academic world and they are not the same by any means um so that is a unique position uh talk about really the differences between uh Academia in science and industry in science um well I'll give you my Impressions coming in into uh academics going to Industry and then coming back so um when I was first in academics I was in academics for about 10 years one of the challenges I found at the time which has changed somewhat was the lack of interdisciplinary activities mhm there was a tendency to stay within your lane in your department in your field and I was more interested in solving problems which meant that would require you to bring in different perspectives and collaborations were just a natural way to do that industry has a different approach industry as they say there is no IE in team mhm and Industry functions usually based on teams mhm MH and so part of the opportunity going to industry was to explore how we could really push some of the collaborative efforts forward the challenge of course is in Academia resources are more limited m in Industry the tendency is to have more resources but the goals tend to be more limited and so some of the freedom that you would like to have you don't typically have I'm not going to say you never have but typically you don't have and so um as I went through several industrial or commercial opportunities I I to be honest with you was felt I was being opportunistic if we wanted to solve a problem and I got to a point where I didn't think we could do it then it was time to look for another place to try to do it um and eventually run out of places and run out of opportunities because again I was interested more on understanding the disease than necessarily developing a drug MH without understanding the disease and I went back to academics thinking that that had evolved and it NE didn't exactly evolve the way I had hoped it had everyone wanted to start their own company and you know the biotech Revolution was taking place the dot Revolution so everyone thought they could be an entrepreneur as well right and and so there's a very different perspective one side tends to have more resource but much more gold directed The Other Side has more freedom or the perception of more freedom but it's much more resource limited yeah and um you know a lot of people don't know but for instance NIH funds a lot of the research at the uh in Academia in the United States and you they have about a $45 billion a year budget overall give or take and uh you know kind of the model at least as I see it and you can correct me here um but a lot of people go work for NIH for 25 years or so um they leave NIH to go work for Pharma many times on some of the similar technology they were working on in NIH right NIH then funds the Pharma companies you know that that they're working at um including Academia as well and it's kind of like a cycle so why while uh Academia is one is different in many ways than uh Pharma um and different than NIH they all really are uh connected together because you need them all to develop yeah I mean in principle you've got an ecosystem mhm and in principle Academia has should have more freedom to explore things the way the funding has evolved over the years though you'll see that it has taken on a bit more of a business orientation M driven by you know what's fundable what's popular um especially as technology comes along what are the new technologies we would like to apply and so that has tended to have an impact on the kinds of research that gets done and the people that going in with looking at new careers coming out of their programs graduate programs postdoctoral programs they have to understand that that is what Academia has started to evolve towards in industry a lot of the kinds of things that are NIH funding which you mentioned it's not the same kind of funding as you have in Academia so you don't have sort of a basic research effort you have more Cooperative research agreements which are contracts that NIH sees that um industry can deliver on because they're used to delivering on contracts that Academia is not is attuned to or trained to be able to deliver I see and you know when I was in Industry we would see this as well um a lot of Industry works very closely with Academia it's a challenge um industry doesn't always they understand they're not always going to get exactly what they would like out of that relationship and Academia frequently looks at it as a source of funding um but not necessarily with the same guidelines and constraints that industry would like them to perceive or understand yeah but NIH can fund both though too right so it's like they but in different mechanisms not the same mechanisms um has that cut down the research cost of the Pharma over the years because I've seen debates about it's not a major not a major input I mean you may have it in a very directed program like a vaccine program because there's a need for a vaccine program but in terms of overall research and development costs farma really um puts the majority of that funding in I see you can look at the numbers on from the Pharma organization M and you're also director of translational medicine for the Pharma foundation too right I I was just until just recently both for translational medicine and for informatics um talk about translational medicine a little bit you're also uh an editor for the Journal of translational medicine um so what is it yeah um Pharma several years ago wanted to set up translational medicine as a program Pharma has a a a philanthropic arm called the Pharma foundation and the goal of that is to promote education it's not tied directly to the fora Institute of Pharma industry so it's not to build a feeder farm so to speak for the Pharma industry but to sponsor research that pharmac sees as a need long term so um informatics had been involved for a long time I started when it used to be called bioinformatics but we expanded it to include clinical informatics and and so on and there was an interest in Crea in translational medicine actually translational medicine and Therapeutics was the original title and so we put together a Pharma put together their board uh a process involving McKenzie doing a survey and then brought together a very large group representing government industry Academia to discuss what a program would look like and um having been in Academia and in industry I had raised the concern they asked me if I would share it to get it started and and I said I would but we needed to I I I I wanted to have certain freedom in doing that and the freedom was let's not take what at that point was the conventional view of translational medicine which was let's take research out of the laboratory and move it into the clinich and the reason was that in while I was in Academia I saw incredibly great science being done MH the concern I had was I didn't see that it necessarily was solving clinical problems and so the concept of that translation was not really Bridging the the valley so to speak and yet the academics felt if they were doing really good researches should convert to clinical utility MH and the you know clinicians and basic scientists don't exactly speak the same language they have different missions yep so when we started the this program I had people from The Gates Foundation from different uh uh NIH or government groups academics and Industry and what we decided to do was say if we're going to do translational medicine and the goal is for research in the academic environment to be able to move to the clinic we have to start understanding first what the clinical problem is M and therefore we're going to start and take not a bench to bedside model but a bedside to bench to bedside model interesting so if we take the problems out of the clinic bring them to the bench to work on then we know that the solutions will have an immediate application yeah and and we've closed that Gap that's interesting so you know um for people listening translational medicine in general would be considered bch to bedside that's like the standard definition and uh basically translating the findings in the lab on the bench right to the bedside uh of the patient through the clinician right and um you know one of the interesting things that that I found uh running running the hospital and being so involved is I think from the outside most people think that doctors are scientists right and scientist um and somehow and that really was the case you know 50 or so years ago was it was more of the case where clinicians would make discoveries and they would notice things and they'd you know really expand upon that but um uh what you're saying is actually a little different I haven't ever uh thought about that way bench to bedside back to bench and um you know I I like hearing that because that's kind of the model that we have set up uh in Mexico where we have um you know our sixf floor cellular lab we have our clinical lab genomics Lab digital pathology uh and the doctors and scientist work area is the same so they're they're working collaboratively to do that there's there's a couple of things to point out I think one is um the majority of the way medicine's practiced right now Physicians have to be operational they're not given enough time because of the the way it works to to Really devote as much time as they might like to devote to understanding the patient and their condition and the patient is putting pressure on the physician to come up with a diagnosis and a treatment and actually the internet doesn't necessarily help because they're coming in armed with all kinds of information and misinformation and the doctor frequently has to sort through that as well so it it's not a perfect match one of the things we've tried to stress is let the doctor be operational but let's find out what is the Strategic IC side not the operational side that they really could benefit from having more effort to be involved in and you know a common common thing would be a easy thing to understand the doctor has to come up with a diagnosis okay the diagnosis gets tied to treatment it gets tied to reimbursement hopefully to improve patient outcome but they don't have time to really study the underlying pinnings of what the diagnosis actually means and so we spend a lot of time exploring how do you make a diagnosis how accurate is that diagnosis does the disease have to be stratified because maybe there's five different subtypes in that diagnosis Y and the doctor doesn't have the tools necessarily to apply it that way unless they have the experience already they've seen it so we try to take that on now why that's important is when I left academics we were trying to do a lot of that but it wasn't sort of the current model and there were a lot of things we were trying to do because this was very early and and it gets to the model that that you have in Mexico we're trying to say instead of trying to take all the D databases and put them together the way people even now try to do let's change this and make it a patient Centric model MH because in essence the patient owns everything they own the tissue they own the disease they own their lifestyle their environmental exposure their culture their social determinance and let's make it patient Centric cuz that really is what the is all about and let's use that as a computational model as well because now we can make the model of all of the data represent the patient MH and we can ask questions about that yes that wasn't the way things were going at the time which was around 2004 in the academic or academic medical community and so that's when I left to run the Windber Center which was the Department of Defense breast cancer Center working with Walter Reed because whereas where I was in academics I kept being told well we can't do that that's not the way we do it when I presented it into the the dod it was well if it makes sense we'll do it m and we don't have those barriers M we don't have those structures what was that like like when when you were able to do that and go from all the barriers to uh being open what what did you accomplish we could ask questions that we we couldn't start to ask before we would see things relationships among data among patient characteristics we would see the most important thing is we could start to better see where there were gaps and conflicts MH because the thing about science is it's constantly evolving MH and it's never going to be perfect it's never you're never going to understand everything but if you can identify where there are gaps and conflicts you have a better idea of where to prioritize your research M and your investment which ones are important where will you get the return on investment so to speak and changing the model gives you more freedom to do that because we weren't going to have to align with what was an NIH review panel going to want to do when everyone on the panel did the same thing right and you're trying to change something yeah give an example uh working uh for the dod um a change that you were able to make that that made a positive difference well the the data model the yeah we we we went from the concept of an EHR electronic health record where you would put all your data in to a personal health record and that started to then so the idea would be uh a patient who would come in to a doctor's office and you would ask do you smoke do you drink how much do you smoke how much do you drink because I'm trying to assess your risk for breast cancer MH but the reality is that patience actually under goes developmental change from the time they're fetus in during pregnancy all the way through their the rest of their life and the breast is actually a perfect example of that there's 10 different stages of breast development just during G ation and exposure to these factors like smoking or which aren't from the fetus of course but maybe the mother and the environment they're going to have different effects at different points in time of that development so risk is not something that is as convenient to model as asking do you smoke and maybe you should stop smoking it's maybe when you were smoking at the time of menarchy that was when risk actually developed but we couldn't ask those questions the same way but we were free to do that kind of modeling and expand upon that with the dod experience yeah so that was uh you said 2004 is U uh I guess the Human Genome Project was already underway uh and um you know that started what in the early 90s I believe it was late 80s early 90s yeah actually when I was with Amo it was before the Human Genome Project started and we were working with Carl wo at the University of Illinois when the pro carot Genome Project and the idea was that would be a model for the Human Genome Project so Amo was involved in that and then Amo of course because it was uh an energy company was involved with the doe the department of energy and the initiation of the Human Genome Project yeah I mean it was a heck of an accomplishment overall I mean if you look at it I what was it like kind of living through that I mean I'm sure you know you had some scientists that were like this is going to immediately make a big difference I say this because uh I think about where we're at with AI right now and um you know maybe well you know it's interesting because I remember I was at a meeting in Waterville Valley which I remembered being one of the very first meetings talking about this and and while he got up and said I'm going to sequence my genome okay and and you know you can see how far we've come since then because not everybody believed it at the time interesting um so we we've come a long way but and you refer to AI we still have the challenge that we can generate a lot more data M than we can extract both information which is clean data where we have context and definitely much slower to create knowledge or clinical utility out of that and and so that's really the challenge we keep developing new and new technology to give us more and more windows so to speak on the problem but they're not necessarily all developed because we know what the question is to ask right they come about because we have the technology that can make the measurement yeah so it's a a bit of looking for your keys under the lampost but now with big data we have a lot of lamposts a lot of people looking for their keys and they may still not find them right well you've been involved in uh understanding Ai and uh well you work with Quantum Computing as well um you know for for a long time so and it's interesting as this AI the last year it's gotten popular with Chad GPT and um but you've been working with this type of stuff for a long time and have understood its limitations but we yeah we started working with AI back in the 80s when I was with Amo and and actually we were we brought it into Amo at a certain level and um we were called into other parts of the organization we were obviously interested in applying in biology and healthc care but we were called into the refinery we were called in to um work with a credit card division because they wanted to see if AI would solve some of their problems but um one of the things we found was we could take problems that seem to have been solved using conventional statistics and we could apply the AI in that case it was neural networks and find more complex relationships that hadn't been seen MH because the statistics would give you a not that the neural Nets were not a form of statistical analysis because they were still corative but they would allow you to look for secondary and tertiary interactions that you wouldn't always find through the other means so but at the same time that's still and to this day is still a challenge cuz we're still interested not in correlation we're interested in causality mhm because if we really want to change medicine and improve health care we have to understand what is the cause of the condition how early can we detect the condition or someone's at risk for the condition if we want to make and have an impact not are these symptoms related to the condition MH yeah and you know one of the things that you've explained to me well is the AI is only as good as the data that's being inputed so if the data is all wrong then you're going to get the wrong answer ultimately yeah there's a lot of emphasis on big data and the challenge is not all data is created equal um there seems to be a bit of a difference between data scientists and data Engineers um I'm working right now with a data scientist and her training didn't expose her to really looking at the what the data represented as much as how do you process the data and use the methodologies and and that's one of the reasons she has started working with me because data that has the same label in different databases doesn't mean it was collected the same way or means the same thing and as a result you have a problem because when you put that together to create Big Data the data is very heterogeneous so I'll give you an example and and actually has it translates to real life so I always use this example the gold standard for understanding kidney function is glomular filtration rate which involves the processing of fluids by the kidney coming in and going out the most correct way to measure that is to put the patient in the hospital control the fluids and have accurate measurements but that's complicated and complex and most Physicians don't want to put the patient in the hospital so what they've done is they've created what they call estimated glomular filtration rate egfr and that's in a lot of patient records and it's computed using some other laboratory measurements creatinin and and Etc but what's not necessarily acknowledged is the fact that there's at least five different equations to compute egfr they were developed for different reasons so they're not directly equivalent they don't give you exactly the same number and two of them have a race factor involved which has been proven and so if you just take different databases and put the egfrs together you don't know where it comes from and it's not annotated MH and so you're putting data together that may not have the same quality or the same basis and and to show you how that really has an effect I was having a conversation with a transplant surgeon who was acknowledging this issue M uh at hyms and two days later CNN had an article a report of an African-American woman who had been on the transplant list for five years but was not given a transplant because they used the wrong equation to compute her risk mhm and once they switched away from one of these race-based equations she sh suddenly was elevated on the list and got a transplant gotcha so she actually was able to be treated appropriately but had a fiveyear delay in that treatment so these things really do matter well yeah and and you know with the AI I there's a lot of applications but uh one of the things that you've taught me is you know a lot of this is based off of people's medical records and the medical records are based off of insurance codes in many ways well some some aspects yeah some aspects and so you get these data points that aren't they don't give you really a clear answer of what's going on and let's say that those those data points give the specific diagnosis but um you know what it might mean under one Insurance claim is different than on another as well and so they're getting right bad data that then you know if you're using the AI for that there's no way it can actually know because it's given the wrong information well so so just just to clarify mhm the clinical data per se has a kind of problem could have a problem like I just outlined right but tends to be numerical and and therefore a bit more reliable sometimes though the data that's recorded is based on setting a threshold that something's positive or negative and then it depends on who's setting the threshold because different guidelines May set the threshold differently MH and that's something that you need to take into consideration the challenge which is primarily in the US is that the majority of the data that's available for analysis is claims data and as you said that's tied to reimbursement it's not necessarily tied to understanding the disease itself and and so I always find it interesting because my my uh genetics groups friends who are always looking for genes associated with a diagnosis and are using a diagnosis code an icd10 code as an example in the US they are using a code that when you talk to the clinicians they'll tell you well we had to put that down because we needed it for reimbursement y it had to justify the test we wanted to do or the treatment we wanted to use it's not necessarily representative of the disease and so that tends to obscure that now outside of the US where you have nationalized healthc care the diagn diagnosis code could be much more reliable M than it is when you're using the claims data that's based off of a reimbursement model I see it it'd be interesting to do a a study comparing same patient claims dat versus their medical records and see what the different outcomes are have you ever seen that done uh I haven't seen that I mean it's definitely something that would be of value yeah absolutely someone must have done that someone had have done it not we figure out how to do it um so uh let's go back a little bit to um translational medicine right so so just to back up to where you were talking before yeah the same transition going to the dod and not having as much of the embedded culture in place is what you have the opportunity at chipsa to do because you're not working with an embedded culture and you can focus more on the problem and bring the pieces together necessary to try to address the problem right and and so that's where a lot of these opportunities can be seated and started yeah no it's it's exciting because we're kind of getting to build this from the ground up right and what most people know us for now uh is stem cells you know the the CPI model but there was chipsa uh well before that and uh we're actually in in the middle of the uh name change translational Advanced medicine Tam Center uh translational Advanced Medical Center I should say Tam Center um and we're really going you know not not dismissing what we did with chipsa because I actually we did a lot of great work but evolving to where it's all science-based you know we were having the discussion yesterday um it's going to be hard for someone to look at what we're doing and say this isn't science if we could have bioethical questions on how to that we need to solve and I'm you know I'm fine having those conversations I think about it a lot I know you think about it a lot as well our whole team does but um you know what do you see as the potential for the lab that we're setting up between the cellular lab and the genomics lab digital pathology all those things uh you know and some of the things we were talking about yesterday those extra let's say 10 steps or 10 other things that we need to figure out in between what do you see as the potential for uh what we're building well the the potential is that starting from scratch as you put it you can start to assemble pieces um based on what the question is that needs to be answered okay and and you can approach it in a novel way not necessarily replicating what someone else has you know I once had um we were working in a pediatric uh uh ncu mhm and uh colleague of mine brought me in cuz we were working in Pediatric ards and there was an infant on the lying on the bed the parents were in the room there were five people looking at this instrument baby was strapped in and the instrument had five dials basically to make it simple and they were turning the dials and seeing what the response was and turning the dials and and we spent about an hour there and we walked out the my colleague was the head of the niku and he said so what do you what did you think of what you saw and I said well from my perspective I just saw a random walk in five dimensions for an hour because they couldn't tell you if they repeated anything okay and I don't know what the coordination was of all the dial you know because I'm not a clinician of course I know that there was experience and and so on that they were looking for and response and a month later I was at a Midwestern Medical Center a big Medical Center and I happened to be talking to the head of training mstp Medical Science training program and I said to him I said you know humans are really really good in two dimensions number of humans are okay in three dimensions four dimensions is very hard five Dimensions is impossible okay so how do you train your Physicians to deal with this because it is a five-dimensional problem MH and he said he said well we only really train him on two dials at a time mhm okay and so I said to him you know so why do you have an instrument with five dials I said is it because the engineer knew that each dial was worth $25,000 and you put five dials on the box and it's now $ 125,000 instead of 50,000 and and and he laughed and I said or is it because the hospital across the road has the same thing I said it's because the hospital across the road has the same thing MH and people are going to expect five dials right exactly right yeah and so part of The Challenge from the way we build models is we treat disease as a process and a chronic disease also has human development under that and the challenge of that of having disease as a process and not a state because we talk about a disease state is that you have to look at how things change over time mhm and that is a big challenge because if a patient comes in with a disease presentation of symptoms it's not ethical to not treat them and just allow them to progress to understand what the diseases I mean you may do it for a small amount of time to see if they've resolved but you're not allowing them to run the full course that way and so we don't know what the real natural course of a disease really looks like how many dimensions in terms of clinical parameters it involves and if you intercede that at different points in time and you're only extracting the data points at that point in time you may diagnose it differently because different diseases may cross at different points in time as they evolve and now what you're trying to do with technology Diagnostics and biomarkers is you have something that can make a measurement MH and that technology and that measurement you're trying to figure out how to align that with that disease process and it may be very different at different stages of that disease process but if you don't know the disease process it's very hard to do that alignment yeah and so you're constantly fighting that battle MH that challenge of getting that proper alignment to be able to really hone in on what we call what is the the disease Vector in high Dimensions how far long that Vector a patient is which is their stage and then how quickly are they progressing M because if they're progressing slowly you manage them differently than if they're progressing rapidly M but if you take those three parameters then what you really have is you have a tensor not even a vector mhm and and I call it sort of the the Heisenberg uncertainty principle of disease CU you can't measure where a patient is on that vector and how rapidly they're progressing at the same time which gets to the physics part right so that's the kind of challenge we have and I don't know that we'll completely resolve it a lot of the new digital monitoring 24-hour monitoring can help give us a much better profile of these changes over time Y and and that's that's one of the big values of how we should start to use that data yes but it's still going to be a very complex marriage of the data with understanding what a disease process is well uh let's talk about that uh for a minute because uh just to give people an idea of um something you're working on uh you know with the underserved communities um uh specifically in pregnancy um what are you working on with that how are you setting up that model you know with the earpiece and and those type of things yeah so um we have a women's health program and it starts with preconception we spend a lot of effort in infinite maternal morbidity and mortality which is a very big area of Health disparities we go into postpartum and developmental ISS isues and then we actually go into per menopause and menopause and postmenopausal disease risk cardiovascular disease and breast cancer but looking at the infinite maternal morbidity and mortality one of the things we focused on are hypertensive disorders of pregnancy which are very common the most common one the most notable one is preclampsia which most commonly results in premature birth the cost INE in premature birth and the result of these hypertensive disorders is somewhere in the first year aggregate in the US over $2 billion wow and of course these can have a long-term impact on the Infant so it's not just that they resolve but they may have developmental issues and so on so we became interested in the idea that um we know with the health disparities there's a much greater risk for preeclampsia and hypertensive disorders say in the African-American Community than the Caucasian Community MH we also know that hypertension in general is more common in the African-American Community than Caucasian Community but it's estimated that in the US half of the entire US population has hypertension MH okay globally it's about 30% but in the US it's about 50% so one of the things we thought we would start to do is see how early we might identify that a woman is at risk for developing hypertensive disorders to try to intercede earlier monitor them earlier and also recognize that while we talk about preeclampsia and this gets to the issue I was talking about before stratifying the condition it's not one condition there's a set of parameters and it's multiple subtypes of preeclampsia that need to be dealt with and they may have different targets for drugs the different response different critical interventions and and so on so we became interested in trying to piece that together in a model the goal was let's see if we can start even preconception and evaluate hypertension there or risk of hypertension and see how that progresses MH that led us to start to look at how is blood pressure actually being measured how is hypertension being monitored and the challenge we realized is that blood pressure actually in the clinic is not even measured according to the guidelines that exist you can talk to almost any clinician or you cardiologist and they'll tell you we no we don't exactly follow the guidelines we do it the way we do it m and then based on what they do they look at a chart and assign whether you have hypertension right okay but the reality is your blood pressure varies over the course of a day normally by about up to 20% mhm and most notably at night there's a dip called the night nighttime dip and that's known to be different in African-American population than Caucasian and there's con there's concerns and studies as to the rate of fall or the rate of rise in the morning that people are greater risk for heart attacks with that rise in blood pressure in the morning but we are taking one measurement typically to represent that entire transition mhm and so what we' started to do is look for ways to actually monitor that entire transition and parameterize it so that we could take features of it how rapidly our changes occurring and then see how those relate or may be predictive of things like preeclampsia as opposed to that single measurement and if that's so can we then look to make a better decision about how to treat those individuals because what's the underlying pathophysiology that causes that feature and then let's get the right drug that attacks that problem because right now there's about 12 classes of any hypertensives and choosing one is not necessarily the straightforward right Physicians will have their favorites and and and so on but you're in a sense titrating the patient yes to see what they respond to and this is trying to change that personalize it and it may change when you get dosed and it may change the order in which you try medications right I think you there's a few things there's there's a few things to unpack um you know specifically uh the the way we measure blood pressure is not really it's it's standardized but it's not standardized uh people don't do it the same way um the other day I was getting an IV and uh they were going to take my blood pressure and I was laying back in the chair I was like no I got to sit up and get my my feet on the ground and then the other nurse like oh yeah you're supposed to do that I'm like well that's a great example I thought about you um that they were going to do it the you know the wrong way uh to where you don't get an accurate um uh view of your blood pressure um the other thing is your idea is you've got a a little earpiece that we're working with a company that developed an airp we don't have the earpiece sorry yeah you're working with a company but but for your for your project uh is to have that earpiece um right uh that the patient wears and it monitors it accurately throughout the day right so we understand uh the answer to that question a lot quicker you know yeah the Ian I mean right now what happens is um a patient could be fitted with the what they call an ambulatory blood pressure monitor which essentially is a cuff that every 30 minutes blows up just the way your cuff normally does The Challenge though especially if you're looking at it overnight is it interrupts your sleep yeah and that's going to cause your blood pressure to change as a result of the intervention to measure your blood pressure and even during the day if you know it's about to trigger you're going to have a response an anticipation so while it's much better than a single measurement it's still not a non-obtrusive way of monitoring and what this potential opportunity is is to look at a a recording that's noninvasive and doesn't have that same it's passive completely passive yeah and you know what's interesting to me I mean that's just one application to right the model that you created and to give people some context uh 2019 we were launching United Cancer Centers uh or raising money for United cancer centers in the United States um you had a model that like to say we weren't ready for yet because I didn't exactly understand it um takes a minute you know I'm not a I'm not a trained scientist like you but um you know we were very much uh we were patient focused at at uccc um we were going to be the first institution to utilize the right to try legislation um and you wanted to take it a step further where we had our own genomics Labs we were doing our own pathology I think uh either you or VJ brought on the digital pathology but um the idea was to get as much information as we possibly could uh from the patient's tumor mhm and then deliver Precision medicine to the best of our ability and build out the labs with UCC now the time uh I knew a little bit about genomics like just very very basic but hadn't really dove in I'm like oh gosh I don't know how how we'd put all that together um but as uh Co happened and that project ended up not happening uh it's like hm there's an opportunity to do that um you know in Mexico and even be able to have more freedom to do more right like uh my question with the with the solid tumor genomics lab is how do we build the best solid T genomics lab on the planet at least for the patient specifically there's other places that'll have our capabilities but is it going to be centered towards what the the information the patient needs either now or in the future right now there's going to be discoveries uh on the tumor that we we can't really take action on but in a year two 5 years from now you know there's that possibility so I like to say that a lot of the idea for what we're doing now came directly from you you know this was uh part of part of your vision for what we were doing and um you know I say that because the earpiece is a is another uh well I said the earpiece the the project that you have with with the earpiece uh and the the pregnant women um is is another part of that M because the model that we're created in Mexico we're going to bring to Nashville um we have the potential to ask questions in a completely different way and focus on these underserved communities to make a huge huge impact and you can do that without a lot of money I mean it doesn't take a ton of money to uh once you understand how to build the labs and run the tests and what tests to run and uh you know how many samples you need before you can run the machine at a coste effective way there's ways that for a small amount of money you can get a lot of meaningful information to patients on the cancer side but also just on the overall uh lifestyle life side um and uh kind of elaborate on on that because I don't think people realize maybe that's what we're talking about in the whole like this is a a an ecosystem that uh you're putting together based on how to solve complex diseases and how to really make make the world a better place or healthier place for for everyone yeah and and I guess there's sort of two parts examples for that um one is as an example there's a lot of emphasis right now on social determinant of health and they're important there's no question about that but I think a challenge that we haven't necessarily adopted or acknowledged is to a certain extent we're looking from the outside in and trying to make an assessment of where certain characteristics features um should be could be and should be you know do the is it an educational problem is it an access problem is a smoking problem things of that nature M we also need to expand upon that which is something we've been doing and say that's from the outside in you need to understand it from the inside out because if you really want to be patient Centric and especially if you're going to be dealing with multiple cultures they don't look at the world the same way you do right and you need to understand how they look at the world because your priorities may not match their priorities and as much as you can explain it they may not accept it and act on it sure and so it needs to be an interface between the two and then even with several cultures at least you need to understand the concept of trust are they going to even trust what you're telling them because simple examples in trust are do they have confidence that you know what you're talking about and do they have confidence that you're looking out for them not for some other purpose yeah absolutely and all of these things come into play that make the problem more complex than even just looking at the social determinance but that's what the real world is about and that's what a real patient is about absolutely and so we need to keep those things in mind on the other hand again with the technology as I said before we take advantage of the technology we can develop and then try to use it to look at disease M I'll give the example I'll use is when I was at VIIs where were working on the her two new test using fluorescent citto hybrization or fish that measured gen copy numbers that was different from the imunohistochemistry test IHC which looks at an antibody response to a protein and yet both eventually were proved for use for looking at what they call her two new positive breast cancer patients which are about 20% of the patients the challenge was they don't give you the same answer necessarily because they're not looking at the same thing and we have to understand that biology is complicated so going from the copy number of the gene which ties to the chromosome to the protein there's multiple steps and Transformations yeah and it may be that those Transformations are the critical thing and we're only looking at the end points and by not looking at the part in the middle we're missing what really is important and the suggestion that that could be a reality is we're no longer looking at her two new positive and her two new negative we're also looking at her to new low MH so we've started to increase the classifications yes and and that shows you at least that science is continuing to evolve but it evolves very slowly because the tendency is to try to come up with classification and put things into boxes yes whereas the real world doesn't really like that yeah and and a couple things I want to actually go back from the copy number to the protein yeah and go through that system but you know what you're talking about with her to um I feel like I have a good amount of experience with I've seen it in action right and um I shared with you a case we had a patient I'll just share with the the listeners had a patient who um uh met athetic cancer um it was a stomach cancer didn't know exactly which type of stomach cancer um they were at a specific Hospital uh friend contacted me she hadn't eaten in you know weeks she was emaciated um he said hey can you help her I said let let me see if uh I'll give it to our doctors and and see what they think got back to him pretty quick and it's like you know she might pass in the hospital within 7 to 10 days you know so we've got to be you got to know that set that expectation um but we'll we'll give it a try well she gets in and um you know the the institution that she was at is a great institution uh it's like a thousand pages of medical records or so it's like a uh and uh we started digging in and noticed that a genomics test had been ordered um you know on the uh uh on the the tumor and it had not been put in the system and so uh we called the next day and they sent it right away it had an actionable marker her to mhm and we'll get into her to what you're talking about low or or regular but um within a week she was eating within two months her cancer was gone not cured but you couldn't see it in images uh couldn't tell it through blood work um and it turned out she had a specific uh cancer that what's called cdh1 positive and uh it's a stomach cancer that you have about a 75% chance of getting the stomach cancer uh if you if you have the cdh1 and they recommend if you have it that you get your stomach removed um and you know which is a big it's a big deal to do but she got it right at the age that most people get it at and um lived a couple years but that's one specifically where um we had a Target that the her two worked well right uh and um there was a good outcome what you're talking about and this is actually the first I've heard this uh was uh her two low right um uh which is different than her two and um 40% of of of patients that receive her two treatment that are her two positive respond the other 60% don't respond um but they might be looking at the wrong marker they might be looking at uh her to when it's her too low and if we would just if we would test for her too low possibly her too low means normally what they do is they have a threshold MH and that means how many cells light up with her to MH that threshold if you're below that you are negative if you're above that you're positive her too low means that that threshold is now a gray Zone mhm so it's not a different marker right well it's a great it's a different uh amount of of her two yes yeah so a different amount of her two right but if they had that different amount um you know could the response be let's say I'm going to throw a number out 100% or whatever uh versus 40% right so so since her two positive is usually about 20% the the issue is and and as you noted only about 40% of the patients actually respond to that the issue is from a systems modeling perspective how do you identify who it's not going to work for and not give it to them so that they can have a treatment that might work for them rather than have a delayed encounter with a drug that isn't going to work their disease continues to progress and the her too or whatever the drug is may actually alter other aspects to make it more difficult when you decide that it's not working it's ch it could be changing their biology it could be changing the biology MH and and that's something that has to be considered from A System's point of view not just who does it work for but who should not get the drug right and so you know that's just one example um of you know blood pressure how they blood pressure is another example of ways that you know the system this doesn't make people bad for doing it this way this is the system but how the system has been created to do things in a certain way and we know there's a better way right I mean there's things that are just blatantly obvious in medicine and science you're like if you did it this way we would actually get a better answer uh to the question that that that that we're asking to the problem a lot of times when I speak with clinicians because I spend most of my time talking to the clinicians what when I present some of these what I consider challenges they go yeah it's a really hard problem yeah but it's not one that I can solve right and I I have to do what I can do and that that is an acceptable and understandable answer but that gives us the idea of where we need to try to contribute yeah to take take away that hard problem and reduce the complexity one of the things that I I enjoy uh about working with you is watching you with the clinicians you know Dr escabedo our medical director has a lot of respect for he's like oh Michael asks the right questions and you know I think Dr escabedo shared with you he's been noting some of the questions that you're asking on how we do uh exams his entire medical career right yeah I mean that's uh what's the so the blood pressure is one um there's another one that you've brought up in the past um is it with breast cancer gosh it's a common a common way that they're uh oh when when they started their uh uh period for the first time right isn't that like a determinant of uh spec maybe something else well we are looking at at menopause uhhuh and and what we're looking at in we're actually initiating a survey because um gynecologists are not well trained in menopause and you can ask any woman age probably 35 or older they don't get the information they really need many of them don't know if they're in perim menopause let alone menopause and that can be 10 years long so one of the things we're doing is we're initiating a survey in the US and in Canada to collect initial data to try to see how women go through per menopause and menopause under common Pathways so that with a limited number of Pathways you can better understand what to expect so as opposed to what a lot of women do is ask their friends you know have you had hot flashes and what did you do as we were talking about before hot flashes may be at a different point in that transition for one woman than they are for another woman and what's important that may you may really need to know is are those hot flashes associated with brain fog are they associated with other symptoms or were other symptoms present before the hot flashes because that may change how you respond or manage that so we actually are initiating a c survey uh to understand collect initial data to look at that and and the kinds of questions that we change to give you an example because we're starting with understanding when menarchy from menarchy all the way through to perim menopause the kinds of questions we're looking at is they're frequently simple statistical questions this is where the AI is applied how many pregnancies did you have but what's really more important than the number of pregnancies you had is how long was the interval between the pregnancies because obviously not speaking from personal experience but a woman who's going through a pregnancy will recognize that she's undergo physiologic change she's also undergone changes she's not aware of that may have to recycle and reset before a second pregnancy comes in or is initiated and those gaps or those periods of time could be very critical in understanding how and what symptoms or how she will progress eventually to menopause so that's where looking at these problems but going to a much higher degree of resolution we think can be helpful yeah to better understand the real physiology yeah no absolutely um and as we're talking about yesterday um from a modeling standpoint of for instance let's say a solid tumor cancer um I mentioned methylation and you kind of chuckled you're like well it could be methylation or could be about 10 other things um you know as in the processes in between but you know going from you know the genomics all the way you know down to um the clinical lab or even the digital pathology um to give people an idea like what we're talking about doing to start with is whole exom sequencing on all tumors um we're also going to be doing a 1021 Gene panel as well which is a bit redundant but this uh it's a it's a good double check as you know how the sequencers work as you mentioned earlier uh you know the 30X means it it does the test 30 times basically to see uh See if there's errors um uh we're also doing um uh cyto kind testing we're also doing digital pathology Optical genome mapping um combining uh the the clinical lab data as well but it could be the in between those things how the process kind of flows through from the genomics to the protein that the answer is and while we'll get a lot of information to start really good information more information that's useful than uh elsewhere we're not able to you know exactly solve this huge complex problem until we set up a lot of other tests in between biology is about processes just like I said disease is about process development is a process m going from what information you have at the genome level to how that eventually translates into risk for disease or actual disease involves a number of different processes some may occur at a DNA level some may at an RNA level some may methylation some may be at the protein level and you don't know if it's one or more than one that is actually involved so what we're talking about is trying to lay out at conceptual level what are the different steps and then map the technologies that are available to see which steps and which parts of that process we're actually collecting data on and try to segment it to get an idea where a certain kind of cancer maybe we're looking at the wrong segment because that's where the technology is it's not that the same technology is necessarily going to be uniformly applicable across every Cancer type or every disease type and so that's one of the things we're trying to lay out and then see if there may be a way to prioritize the next technology we try to develop the next set of measurements or combining existing methodologies to give us other answers that we haven't currently pulled out yet right exactly uh you know it's it's interesting looking at this you know not as a science scientist uh for me like coming in saying okay well it's the first hard difficult part is figuring out all the different equipment that you need like from a from a basic level it's like okay what equipment do you need to do this test who do you need to hire to run that equipment because it's Advanced um what tests do we want to run you know we've got this great lab we're setting up but I'm still learning constantly like oh you we're talking about yesterday we got to see if we need to buy a new machine for this or um we might have to look to the outside to get a test developed or still it still always should start with what's the question you're trying to answer right and if the question is are we looking for protein markers on the surface of a tumor that's a different question from asking how early could we have determine that someone was at risk for cancer right and so different questions require different parts of that entire process right to be the Focus right or what's driving the cancer right now right you know um it's pretty interesting but we're going to have a lot of a lot of capabilities and I think that this model it can easily be trans uh transferred to the US I mean we'll be doing it here yeah it's not not unique in any way no it's um it's uh in many ways common sense I mean it's a common sense approach to trying to solve a complex problem yeah yeah well you and I discuss it's it's basically what I call applying critical thinking to a complicated problem right it's not we're trying to take a complicated problem apart look at the root cause elements of that and then see which of those we can address or we can't address address and what that is going to mean about biasing the results we get and interpreting those results and then using them in the clinic or using them to develop drugs or other treatments yeah no absolutely and um you know from what people are used to let's just say uh right now in the genomics field that's available to patients you know you can do uh whole genome sequencing you can do whole exome sequencing which is the coding regions the whole genome sequencing obviously is the entire genome uh the whole exom is the coding regions of the genome um the you can do targeted panels which are the well maybe you should explain to the audience the exom are the parts of the DNA that code for proteins yes the proteins actually do the work mhm the exomes represent only couple of percent of the entire genome MH so when you're doing a genome sequence only maybe 1 to 2% of that is actually converted into proteins the rest of it we don't actually understand yes so that the exom is the part that we uh under somewhat understand uh the rest of the genome while you might not want to sequence the whole thing I it's a lot of data and um you know it's but we the bottom line is we don't know what we don't we don't know yeah um but uh the exome is actionable in some ways we know we know something about it um and so yeah you go from the whole genome to the exome to the targeted type panels um which are good too because they're the ones that we know a lot about we know that maybe they cause X Y or Z as well and um so in doing that the test that you can and get right now let's say in the US you send your blood work in um you get a whole exome test or a whole genome test um you're going to or a targeted Gene Panel test um it's going to take four to six weeks to get the information back and then you'll get a variety of of reports um and those are available readily right now the the reports though it's pretty interesting because you know there's a lot of bioethical debates about the information that you get um people have to be aware we actually the article that you just gave me um was a good one you have to be aware that okay you do this uh uh whole exome sequence and you get a report that says you're at a high risk for heart disease all of a sudden you might not be able to get insurance on um you know on your life insurance and you know should it say necessarily High risk or medium risk and who who makes the determinance uh of what is higher medium and I'm just using that as one example there's a lot of qu a lot of questions that need to go into um that you need to be aware of when doing you know these tests and things to to consider as well well the it it's unfortunately even more complicated yeah because the changes that you see that you're now associating with that risk as more research gets done sometimes it flips their significance so something that you assume to be at risk cause of risk is not associated with that risk because there's actually two things and if you don't have the second the first doesn't do it on its own so there may be two different markers in your genome that are needed and that changes the whole profile yes so that presents a challenge so that's why you want a profession to give you the interpretation not an online report report so to speak yes and that doesn't mean the companies are bad but it you know it's it needs to be explained and even you know uh we have a lot of discussion at the hospital about you know what is the how how to explain this to patients you know as we roll it out what's the right way what's the way to not worry somebody too much um but let them know when there's an issue and when is that and you know it's a Well the whole concept of risk is very difficult to explain yes if you have a 20% risk versus a 40% risk are you actually going to do anything very different and it depends on what what is producing the risk sure yeah absolutely and and and if it's explained in the right way you know I like to think that and this is one of the things that got us as far at the hospital was we're very open with the patients and we let them know um what the what the risk is what it's not and give it into give it to them in a way that they can consume we build a lot of trust with the patients you know and you know with something like cancer in the past it's a lot of very intense emotional type things but I mean there is a way to do it I'm feeling pretty good when we roll this out because you know it's I feel like we've dealt with it on the most difficult end and if we're focused on doing the right thing and explaining in the right way yeah maybe we can you know be the best at even that part yeah but but again and and this is what I tell people they saw with covid you're actually the part of the challenge with Co I think is people were actually seeing the scientific process MH as it occurred Y which is why the information kept changing mhm because you needed data and you needed to process the data not all the data was good the theories changed and then you entered into other Realms beyond the science that had impact yeah that further complicated it but just the underlying science is going to evolve Y no it's absolutely it's absolutely going going to evolve and and a lot of the a lot of what we experienced in Co to me was like a an ethics of how you present the information to the public debate where people you know uh didn't really know that they're like if though this is a Hardline you know answer do you remember uh uh Joselyn she had her wedding it was a all vaccinated wedding in April of 20121 and um this is when in March they were on TV saying uh if you get the vaccine you can't get covid if you get the vaccine you can't spread covid that was the messaging that was going out so her fully vaccinated wedding about I don't know quarter of the people half the people got Co a good amount and she sequenced it and saw was the Delta variant you know breaking through but that information didn't get to the public until you know July or August and you know it was a it is an ethical debate on what people should know how much information how you give that information publicly and unfortunately it turned very political as well you know you had political debates out there and you know we were talking about boxes yesterday you're in this box or this box it's like I don't I'm not in either box Actually I don't even think like that but this is how I feel on this situation this is how I feel on that situation and that's a lot of my experience of what science is it's figuring out how to explain it in an honest way that you know ultimately meets the [Music] um meets the proper amount of Ethics in society but doesn't um you know doesn't go go overboard or or or too little you know yeah so um yeah what do you think about uh kind of moving forward with uh everything else you've got going on with ipq like what's what's the what's the plan uh for the next um I know you're big into strategic plans what's your strategic plan uh for the next few years we're you know we're we're continuing the development of the women's health program and we're working with um underserved populations we're working with um uh African-American population we're working with the Native American population we even work at an Aboriginal population we work internationally and so part of the goal then is to use it um not only to improve Women's Health in general but to also start to reduce some of these disparities MH and um you know if we can contribute to those kinds of things I think we'll have accomplished our goal no absolutely that's we we don't promise to solve all the problems we promise to help make them transparent so people can make better more informed decisions and hopefully not confuse them but let them understand what the reality is and use that also to drive ancillary research like with the hypertension yep well actually you just said informed decision and what you just explained was what I was trying to say too you know that's exactly it yeah that is that is it how do you how do you allow patients to make a better informed decision and um you know do it in in the right way and uh and ask the right questions and yeah absolutely and ask the right questions and look outside the box and you know you always said that you look at the abstract of the the question um and uh you know from from wide different angles um to solve these complex problems and uh I think by doing that you've simplified things you know we're talking about we say it's Common Sense taking a complex problem taking a common sense approach but you know it it really does in a way take your thinking outside in a different way yeah to use critical thinking to allow that Common Sense approach yeah yeah yeah I I think we we call it systems thinking we call it critical thinking but it's it's understanding that what we frequently are looking at as a problem is really embedded in a much bigger system and we have to know how much of that system is it going to impact either the solution we try to put in place or what is the source of the problem itself um and just to go back to the AI because it's such a hot topic um you know you have the experience of you know you said in the while you were with Amo you were applying AI um you're at a lot of the AI conferences as well um and you see the booths and and those type of things where do you think we are right now with AI uh not just in health but as a society um I think we can differentiate AI maybe a little differently than we tend to because we talk about AI in healthcare there are certain aspects a AI as we're talking about it in general because it's a broad field is can be very useful but it also can be very much abused and and that's what concerns me it's abused because people get very excited it's a new technology they would like it to solve very complicated problems but a lot of them don't know what the question is that they really want it to solve and that's where the challenges it's proven to be very useful in operational issues um streamlining certain processes um things where correlation can be very appropriate but I find it still a challenge to see that it's addressing the more complicated problems M and I'm torn between whether it can going forward or what we're encountering is with all the hype people are using it inappropriately or for inappropriate things applying it out of excitement MH and sort of diverting us from going down the path that will give us the most success interesting that's the concern I have what is uh what's the difference between Ai and like Quantum Computing oh Quantum Computing is is it's sort of like um it's a mechanism that you use to allow much broader AI to be applied bigger computers faster computers Quantum Computing is really um more of the technology that you can use to create the to create the AI you do the AI using Quantum Computing I see and you worked with um Quantum Computing at least on the idea of it for a long time right yeah what was that what was the company you were working with over co uh Quantum Computing Inc I see what what do they do um they basically they're doing a simulation using algorithms that will be applicable to a Quantum Computing environment right now they've moved more into the hardware side mhm from from the um software side but at the time we were doing we we were taking advantage of the fact that we could build very very large parameter spaces gotcha so Quantum Computing is the Computing that uh the AI uses to spit out the answer AI can use it yeah AI can use it what what do you think the potential with the quantum Computing is again you know right now we're still challenged with the building the real quantum computers um it still has to be driven by what are the problems we need to solve it it works very well in security areas in healthc care it still needs to be defined how it's going to be applied um we actually we looking during covid at using it to build very complex systems to try to optimize vaccine distribution mhm we didn't get to do it but it it was the kind of thing where we felt if we built a system with high enough complexity we could almost use the quantum computer to dial in changes and get very rapid feedback as to how to change distribution routes where things were needed and shift things in a very Dynamic manner so so overall with the AI and Quantum Computing am I right to say you're not sold on it being what uh a lot of people think it is today I think a lot of PE you know um the joke I always tell people is I'm afraid that some people who are applying AI don't know how to spell it okay yeah um and that you know it's accessible mhm and people should play with it but I don't know that they recognize what it can and can't do adequately yeah I think it's it's still in what the Gartner calls the hype phase the hype phase have you ever seen the movie Idiocracy no I just had a thought like I wonder if like we're programming this AI to be dumb and you know I I forget how far in the future that show was or that movie was but if we became like these Dumber beings because of yeah programming the AI wrong to to think for us and yeah and I see some of that happening yeah I mean chat GPT I mean I enjoy it there's it's it's very useful in many ways but I'll set it up with questions because I I feel like I know part of its bias and well yeah and when we could we could spend another hour just on that but you know when you become dependent on prompts mhm then by definition you're biasing what you're going to get out whether you recognize that or not right yeah so it's a big blind spot in the AI per se yeah yeah that's that's what I found um I had a friend actually uh uh did a podcast with him that barred the uh yeah it said that he should be put to death which is kind of scary for him he's like yeah you know based on and it made up something that he said that he never said used a quote from him and uh it's pretty you know that's if people believe that that's you know right one I mean there's so I love it don't get me wrong it's fun I use it again I I I strongly believe it has very powerful appropriate applications the challenges people go beyond those boundaries and don't understand that they've gone beyond what its real capabilities are yeah absolutely it can do everything it can solve every problem but well and I I I fear that a lot of society is looking for a lot of things that for technology to address and you know I the other thing I always tell people technology is the instantiation of a solution to yesterday's problem and doesn't know what today's problem is that's really good that's really good anything else Michael I guess we can leave it with that quote but uh you got anything else in closing that you'd like to say no I think that one sort of sums it up all right Michael leben thanks so much for coming on it's always a pleasure talking to you thank you again right