Patrick Short  0:03  
Hi, everyone, and welcome to the genetics Podcast. I'm really excited to be here today   with professor, Sir Peter Donnelly, the CEO of genomics PLC and emeritus professor of statistical Science at the University of Oxford. And actually, Peter, you might not even know this. But I spent a summer at Oxford University in 2011 2012. And I worked at Peter Cook's lab cell biology. And I learned during that summer that I was very bad at cell biology, but I was actually pretty good at computers. I have a math background like you. And when I was looking at PhD programmes for a couple years later, I actually asked Peter Cook his recommendation for people that were not in the lab, and we're a little bit more on the mathematical and computer science side of things. And he gave me your name. And so when it came to genetics and big data, Peter Donnelly  is the man. So I actually had I looked back at my email, I had a brief email exchange with your PA and you and Chris Spencer back in 2013, because I was thinking about applying to the Rhodes Scholarship and I went to work in your research group and I don't I don't even know if it if it necessarily crossed your desk. But as fate would have it, I didn't get the scholarship and went to Cambridge instead and worked with Matt Harrelson. Very happy there. But there is a parallel universe actually where I if I did that interview better if I wrote a better personal statement or something I might have actually been in your research group. And as I was preparing for the podcast, I actually went back in my email and read that and I thought it was funny because it's a small world as well, because Chris Spencer was one of your co founders at genomics PLC right along with Gil McVean and Gordon lunches the roundabout way of saying how how did that come together? The idea for genomic PLC, what got you all thinking about starting a company you are you've obviously had a very significant academic research career. And what caused you to make this pretty was pretty big change? I think at the time, 

Peter Donnelly  1:41  
Yeah, well, first of all, thanks for having me. And secondly, I was sorry about the name is actually it's not you probably remember, not the only time we've tried to encourage you to work in our group. But it's been great Patrick to watch you go from strength to strength in a in your own career. So genomics PLC, where did it come from? We had been and we're very lucky to have been at the centre of have an extraordinary period of 10 or 15 years in genetics. So they've kind of just the early genome wide association studies that my project before that which I was lucky enough to be centrally involved in, it really was an amazing time. And and we're learning so much about genetics and its role in all the common human diseases and in fact, many more human traits. And all of us. And I would certainly do this in every paper, we wrote, certainly in the letter to the editor, and in in our talks. And absolutely, in the beginning. And the end, in the middle of our grant application, we talked about the impact that all of these discoveries would have on patients on on healthcare. And you know, I loved the academic stuff I was doing, it was fantastic part of my life, but it sort of suddenly dawned on me that we weren't actually having much impact on patients on health care. And I and Christian Gillen hurt him reflected a bit on that. And it's, you know, the academic incentives in the academic world, just underlying towards that it's about doing the research, writing the papers, getting the grants, or great. And it's sort of tempting, from the academic piece to think of, of the what we'd call translation getting into the healthcare as a sort of relatively straightforward thing to do next, but but were affected a bit. It became clear, it wasn't straightforward. And that was the driver for founding genetics, we wanted to do two things, we wanted to have a company that was still, we hoped, and I think, I think other people would say, we've succeeded, but we hoped was still world leading in terms of the science and the research. But add to that all the pieces, we needed to work out how to get this into healthcare, and to actually make a difference in healthcare. So that was the that was the drive behind founding a company. It was about making sure that discoveries that we've been lucky enough to be a part of, but but many, many others in the community were driving forward could get from the research papers into actually making a difference for patients. 

Patrick Short  3:49  
 Yeah, and I've seen you present a number of different times. And one of the things I really love about your presentation is the framing you give about your mission to understand the human wiring diagram. I'd love if you could maybe just explain this and some of the different outputs or strands of focus, whether it's genomic prevention or understanding drug discovery. And I love if you could just explain to everybody this framework of the human wiring diagram and what that what that means and how we understand it. 

Peter Donnelly  4:16  
Yeah, love to. So there's a depressing amount we don't know about human biology, a ridiculous amount we don't know about how cells work and how things are put together. One thing we do know is that all of our cells have an instruction manual. It's our DNA or our genome, somehow rubber, they make sense of that instruction manual. And they do a shitload of things. They make proteins, the proteins interact with each other to form complexes do other things. And in almost in a Presto way, we have organisms. And in the particular case, we have humans, and somehow all of the stuff that makes that happen is encoded in the DNA. It'd be great to know that and to try and work out some of the pieces, at least some of the pieces of that. And so one of the framings and as you've talked about it so you

diagram that helps us think about things is one thing we learned from that is genetics is central, you know, everything starts with the DNA, and then follows. So one opportunity we have in trying to untangle some of what's going on, is by looking at what happens when the DNA is changed. One of my colleagues used to use an analogy that said, you know, supposing we were trying to work out how a car was made, and you're given the instruction manual in a really bizarre language that you couldn't read. But then you didn't just have one instruction manual that had many, many copies of the instruction manual, and each one had a slight change. And you could work out how the resulting car made from the change instruction manual was a bit different than you might begin to piece together the instructions. And that's broadly, the theme that underpins a lot of our work in genomics, the company it's using and understanding those changes between people in our genome in our DNA code, and their consequences to learn things that are useful for health and health care. And, and genetics really does play a central role. It's one good thing about it, as I said, everything starts in the DNA, causality can only go in one direction, it has to be that the DNA change causes something. Whereas almost all the rest of biology, you know, someone's sick, and you notice a change in abundance of a protein. Or if you notice a change in gene expression in the sick people, you don't know whether the physiological change is causing the disease, or whether when you develop the disease that causes the physiological change. But if you notice the genetic change has the property that it makes people more likely to get sick, it has to go in that direction. It's not that when they're sick, their DNA changes. So there's something very special about genetics in helping us understand biology. And that's the, that's the framing. And we often talk about using this and the patterns of change to figure out more about biology as untangling the human wiring diagram.

Patrick Short  6:45  
Let's talk about polygenic risk scores and Choma credit scores. Because this scenario, you all have done an incredible amount of foundational research, but you're also translating it into the clinic, I listened to your episode on the bio eats world podcast, which I recommend to anyone. And I think it's absolutely worth the lesson. One of the things that I really liked about it was you and Vinita, who actually is a previous guests in the podcast, had a really great analogy about putting sand or pebbles or occasionally a brick or something larger into a pillowcase to describe how genetic risk scores work. I was wondering if you could talk through this analogy for anyone who's not familiar with the concept on on, on what this is and why it's so important following on from this point you just made about DNA having a central and causal role in changes in human health. Sure,

Peter Donnelly  7:33  
so the basic idea is the following. We've known for many, many years for each of the common diseases, and all the common cancers, so heart disease, or diabetes, or schizophrenia, or breast cancer, or prostate cancer, bowel cancer, we've known that genetics is a substantial part of the differences between individuals and how likely they are to get the disease. It's for some diseases, actually, it's the main factor. For many of the cancers, for example, to others, it's part of the story. And other aspects of our lifestyle and our diet are also important. It's only relatively recently that we've learned more about the nature of that genetic basis. So for heart disease, for example, it's not that there is one gene for heart disease is not there are two genes for heart disease, we now know, there are many, many individual places in our DNA, so individual letters in our DNA, which contribute to someone's risk of heart disease. And in fact, you know, there's something like a million of them. But individually, they have tiny effects, that maybe that if you have an A rather than a to displace on a certain chromosome, your risk of heart disease goes up by half a percent. And somewhere else, if you have a T rather than a G, it goes up by a quarter of a percent, and so on. So you don't worry about the individual positions. But actually, now we can identify those and aggregate the effect. So the analogy and Vinita deserves credit for this. But the analogy is, so we now know which positions matter, because we've done very large studies. So you might imagine, at least conceptually, within a person's genome, your genome, for example, we could start at one end of your chromosome one, and we could walk along the genome. And each time we get to one of these positions, maybe it's it's lit up or something, we can measure what you have at that position, and it'll either increase your risk a bit or decrease your risks. And so if it increases your risk a bit, we might put a little pebble into a sack recurring, and then we walk along a little bit later, we'll get to the next position that matters, it'll also have a tiny effect, maybe it increases your risk by a bit less. So we put in, you know, a small handful of set. And then we go on to the next one, maybe at the next one that matters. You have DNA letters that decrease your risk a bit. So we might take out some of the sand and so on. So we sort of conceptually walk along the genome. At lots of these places, there are a million places we'll have looked at by the time we get to the end of the genome, and each one we're either putting in a little bit of sand or some small pebbles or taking them out. And at the end, when I've done that with your genome, I'll have a sack which I'm sort of carrying over my back which will have a certain weight and that weight for you will be the accumulated impact of all of those physicians. And if I did it for someone else's genome, if I did it for my genome, I'd be putting sand in and taking the sand out. And I get to the end and I have a different weight in my sack. And if your sacks heavier than mine, that means that your overall genetic predisposition of heart disease is more than mine is. And that's the idea behind the polygenic risk score, you know, within a person, we can now measure actually pretty cheaply, we can measure the million positions with medical heart disease, and then we have algorithms which combine their effects to get an overall score, you can do the same for other diseases for breast cancer. For women, for example, there'll be a different set of positions, there'll be about a million of those that matter as well. And again, you can combine those within the person. So actually, you've got lots of sacks, you've got one sack for heart disease, you got one sector diabetes, you've got one for schizophrenia, you've got one for bowel cancer, one for prostate cancer, if you're a woman, you'd have one for breast cancer, and so on. And for each of us, they're sort of telling us what our overall genetic predisposition is, obviously, if we compare between different people, some will have heavier sex than others, the ones with heavier sex are the ones who have rather more of these variants that increase their risk. So we call that sort of collective impact the polygenic risk score, so people with higher scores. And what we've been able to do in the last few years is to measure the impact of this genetic component on disease outcomes. And so for example, without we probably, as you said, we put a lot of effort in our company, to finding really good ways of measuring these or capturing these, but with our scores for heart disease, for example, if a man has their score in the top 3%, their lifetime risk of heart disease is almost 45%. If a man has score in the bottom 3%, his lifetime risk of heart disease is under 5%. So it can have a big impact on risk, it has an impact on risk for women as well. But their overall rate of heart disease is lower than for breast cancer, for example. So their breakdown is a bit complicated, because there are genes that record genes where if you have a mutation, they're very rare. But if a woman has mutation has a big impact on a risk, but ignoring that the polygenic risk also aggregating all of these individually small effects variants, women in the top few percent of their polygenic risk or have a lifetime risk of breast cancer of 30%. Those in the bottom few percent have a lifetime risk of two or 3%. Really big differences. And if you look at their risk across their life, the women with the high scores, they start being an increasing risk of breast cancer in their early 40s. Whereas a typical woman doesn't have much risk before she's about 50.

Patrick Short  12:23  
And how, what are the steps needed to actually get this into the healthcare system into the hands of patients? I know you all are running tests in the north of England at the moment, maybe you could talk through how you get from this observation across many, many datasets that we all see and trust and believe to actually something that that could be deployed into the healthcare system.

Peter Donnelly  12:42  
Yeah, great question. And something we spend a lot of time thinking about and working about. That's the piece that you know, the getting from the research into healthcare, which is hard. So I think genetics will make a difference in a number of different ways. The first and most obvious is about disease prevention. So if we think about healthcare, in general, in an ideal world, we'd prevent all disease there, I can come back and talk about them in a minute. But there if we don't prevent the disease altogether, then the next thing we need to do is make sure our patient gets the right treatment and genetics and these ideas can play a role there. And the next thing jokes can make a difference with and it's another focus of ours is actually finding better and more effective and new treatments. But but it naturally starts with prevention. Many people have argued that the key to solving problems with healthcare systems is to get better at prevention, healthcare systems consume a huge amount of resource. You know, we spend a lot of money on health care, particularly in developed countries. In the UK, the Secretary of State for Health, the cabinet minister responsible for health said in a speech earlier this week that the total government spending on health and related things like social care is 44% of government spending. Well, it's huge. And he was arguing that and the National Health Service in the UK, have had this as a central part of their plan for a long time, as have many other health systems, how do we tackle the increasing cost of health care, we should tackle it by as much as possible preventing disease. So that's where we start with with polygenic risk scores, because they tell us about someone's risk of disease when they're still well. And so we can do that. And healthcare systems already have pathways for either preventing disease, or in some cases, treatments that will put it off or stop it or screening to catch it. Early healthcare systems are already good at this stuff, and they already do it. So what we can do with polygenic risk scores, and as what is now called genomic prevention is a whole new use of genetics and healthcare. You can imagine taking people in their middle age, when they're still healthy, you could get genetic information from them, which we can now do pretty cheaply. You could calculate their genetic competitive risk, the polygenic risk score from those diseases, you then want to combine that with other information about the individual to get what we call an Integrated Risk Score. And we can do that we can identify people who are at high risk for particular diseases. You know, imagine if we did this across a bunch of people in the population. There'd be some who are at high risk for heart disease, some who are at high risk for breast cancer, some who are at high risk for us process. Now, those are people who are at high risk, it's just that they don't know it, and the health system doesn't know. So I often describe it as being able to identify people who are invisible to the system that the system can do something about that they know about. And then you have to figure out disease by disease, or what do you do. In the case of breast cancer in the UK, we offer screening to women by mammograms at age 50. Well, some of these women who are at higher risk because of their polygenic risk score, have the same risk in their early 40s, and the typical woman in their 50s. And they should probably be offered screening earlier. So it's about the way we think about it. We want to improve prevention, by identifying the right people and getting them into existing bits of the healthcare system. And actually, people often think, well, prevention costs money, actually, when you think about it, if the right people are going into the screening programmes, those programmes become more efficient. If the right people are getting put on early treatment or intervention programmes, they become more efficient, you sort of get more value, because you're treating people who will benefit more. So actually, it increases efficiency, and it and it stops disease. And we also spend a lot of time quantifying that to help health systems see it. So it

Patrick Short  16:02  
seems to be there's there's two strands, you've got to show in practice that it improves outcomes. But then you've also got to show that it's cost effective. Are you running those in series or in parallel? So when you run a trial, are you measuring both? are we catching and screening people earlier? And is it saving cost the same time are you are you going to first prove that it you know, it seems to help people and then run a subsequent study to prove that it saves money.

Peter Donnelly  16:25  
It's a mixture of the two and actually the first part showing that you can that these things actually make a difference, we can get a long way there with existing resources. They're very large, I'm sure they've been discussed on on your podcast before, they're very large resources like UK Biobank where there's genetic information available and healthcare information on large numbers of people. Those are real people in in this case, in the UK healthcare system, we can look at the impact of polygenic risk scores there. And we can we and in fact, other groups can do quite a lot of the heavy lifting of showing actually, this really does identify people who are at higher risk for disease. In take us living example, healthcare systems. It's true in the UK, it's true in the US, they already try and identify people at high risk for heart disease, for example, they do it with algorithms that combine the person's age, their sex, their blood pressure, their BMI, their cholesterol level, and so on. They use slightly different algorithms in different countries. Well, genetics can add to that, surprisingly, the genetic component rest apologetic, rescore is independent of the current clinical risk tools, which means it's new information that you're capturing. So you know, we can, we can and have shown to others that if you add this to the existing risk tools, you do a better job of picking the people who are going to go on to develop heart disease, we've done that we've done that in UK Biobank, we've actually done it in a number of other resources. So we've we've validated those scores for people from a range of different ancestries. So some of the work on showing that making makes a difference can be done before you do a trial. And then there are other things you need to check in a trial, you know, we would say that there are women who are at higher risk, you should be having mammograms earlier in their lives. So in a trial, you can do that. And you can, you can invite them to mammograms, and you can see whether those mammograms are more effective and find the cancers that that the existing data would predict. So it's on the showing the outcomes, we can get some way along with existing resources. And then you need to do trials. And then the health economics piece is absolutely done in parallel, you want to do get the information you need to do the modelling, that will show that over larger scales, and over larger times, the savings will be this or, or that. And actually, the third piece is really important in trials is actually just showing how this works and how it fits into healthcare systems. Example I talked about in heart disease of adding genetics into the clinical risk tools. That's something that in the UK and in most other countries has done in primary care in the UK primary care physicians, they're called general practitioners here. They're incredibly busy anyway, and they're particularly busy in the current world. So the third and really important thing to show is that this fits in naturally with current workflows, and how the healthcare system that sort of practical stuff about how the healthcare system works. And as I think you mentioned, we're already doing a trial of that in the UK, in the Northeast of England, we're about halfway through, we've got about four or 500 patients   and for whom we're using genetics to improve their healthcare. We'll finish the trial in a few months, and really looking forward to find out the results of that. We're planning other studies to get to get more of the information you've described.

Patrick Short  19:27  
And in that particular study, how many different diseases are you looking at cross because last I heard on the bio eats world podcast that I think you heard on 45 that your model could do at the time. I'm sure there's many more that are in progress or at that point, but how many of you actually testing and in the real world at this point,

Peter Donnelly  19:44  
that study, which I think is the first or one of the first that's ever added polygenic risk scores into clinical care is just looking at at cardiovascular disease prevention, and then were close to larger studies where you would simultaneously look at maybe six Some of the really common diseases. What's significant about this approach is that it deals with the genetic component of all of the common diseases. Yes, something like 70% of healthcare costs occur with the common chronic human diseases and the common cancers. And that's why preventing them offers so much. I mean, that sounds like a financial point, which in part it is. But actually, it's a point for the patients detecting disease early or stopping it from happening. I mean, of course, it saves the healthcare system money, but it's much, much better for the individuals, it improves outcomes, getting the right people into the right screening programmes is good for the individuals as well as good for the efficiency of the system. So in time, I think this prevention approach will be done for multiple diseases. And one way of thinking about it is across, you know, across 25, or 30 diseases, our polygenic risk score for any one disease is sort of independent of that score for any of the others. So across 25 or 30, diseases, most of them, most of us will be at high risk for something Yes. And we don't know what it is. And this gives us and critically in the way we think about it gives the healthcare system a way of knowing what that disease is. And in many cases, they've got existing pathways in time, I think a new pathways will develop be developed. But it's not, we don't have to develop a new way of dealing with it. It's just identifying the people who can benefit from the existing screening programmes, existing prevention programmes and the early treatment programmes.

Patrick Short  21:18  
Yeah, and that's one of the most, I think, exciting, but also challenging aspects of this whole idea of genomic prevention is you have potentially this approach that could give 25 3040 different risk scores, and almost everyone would have at least one, but the healthcare system really hasn't been built around that paradigm. And so it becomes challenging to map out those pathways figure out how to pay for it, because the system's not used to seeing that kind of comprehensive risk scoring all at the same time, right is that one of the big challenges of it seems like it's hard to leapfrog to a future where everyone at age 40 gets a genetic test and a readout of their, of what they're at high risk for and then moved into the appropriate pathway, we're feels to me like the system isn't quite set up for that yet. But the technology is pretty close to where you could, you could probably deliver that.

Peter Donnelly  22:08  
That's right, I think and it's about baby steps first, and then walking and then increasing speed, I think the natural place to start what we have started with heart disease and cardiovascular disease, because the system already measures risk there. And it already knows what to do for people who are at high risk. And the only change is that instead of combining age, sex, blood pressure, cholesterol, BMI, and so on, the algorithms combine age, sex, blood pressure, cholesterol, BMI, and prs. So it's a very slight change, it's relatively easy for the doctor to explain to the patient, it's not a scary thing, I think for the patient, although we're going to find that out in the trial we're doing. So that's a natural place to start, then the thing, I think the second step is to pick a number of the really common diseases where health systems already have screening programmes in place. And they already gate those screening programmes with estimates of risk. Now, in many cases, they're very high level estimates of risk, like someone's age. So in the UK, for breast cancer screening, as I mentioned, women are offered screening when they turn 50. And that's sort of 50 as a surrogate for some level of disease risk, when you know, the screening programmes in the UK thought it would be a natural start, the way we would say is we can just be a bit more sophisticated about risk instead of saying everyone's the same. And we'll wait till they get to 50, we can say we can be much more personalised and we can identify the people who should have screening before 50. And the UK already has guidelines about women who are at higher levels of risk, and then as an MCI levels of risk, and already recommends that they get screened in their 40s. Now, because of the polygenic competitive risk, 20% of women fall in that category, where UK guidelines say they should be screened. So as I said, the first step is something where it's done routinely anyway, I think the next step is a set of diseases where the screening programmes and the pathways are already there. They're already gated in terms of risk, but usually not very sophisticated assessments of risk, we can do that more effectively. And actually, in that context, there's an equity piece, if you're saying someone at this level of risk, because I happen to be 60, or 50, or whatever the the age for the programme is get screening. If there's someone who's at the same level of risk, it's actually it's hard from an equity point of view to say, no, if your risk is at this level, because of your age, we'll screen you or if your risk is at this level, because you've got a rare genetic mutation, we're screening. But if your risk is at exactly the same level, because you have a million common variants, which collectively give the same risk, we're not going to do it. So I think there's a there's a very strong equity piece as well, which I think is helpful in just thinking about this. It's about it's about being more sophisticated in the way we use risk to get the right people into health pathways. And the natural place to start with is the really common diseases where the pathways are already there, and they're well oiled and they're effective.

Patrick Short  24:55  
It's precision medicine in the clearest sense of the word right using using data Absolutely to to better move people through towards health. How do you you have done a lot of really great work on unpicking some of the big challenges around making sure these scores are as effective as possible. And people have different ancestries. Could you talk a little bit about what the challenges are there and and some of the work you're doing, because that's also such a fundamental piece towards getting it into, into having an impact on patients?

Peter Donnelly  25:23  
Yeah, it's really, really important. High level, I think these sorts of approaches have enormous potential to reduce health disparities, and to reduce inequalities in health care, because by definition, you're being systematic, by definition, you're doing a sort of blind approach to assessing risk and capturing people, they don't have to be already engaged in the healthcare system, in many ways, if you're doing sort of more systematically, so the potential is huge for tackling health disparities, which is, I think, really critical issue for us, you know, as a society, it's true in the UK, it's true in the US, it's true in most developed countries, certain groups of people have less exposure to an engagement with healthcare systems and have worse health outcomes. So so it's a major major issue with polygenic risk scores polygenic risk scores as prediction tools tend to be more powerful in individuals of the same genetic ancestry as the ones in the original studies that we use to derive the scores. And for reasons which regrettable, very regrettable, I think, a large proportion of the early genetic studies were done in people of European ancestry. So there's just less data on people of African ancestry or different types of Asian ancestry, and so on. So that means if you just use the data that's available on those groups, and develop a polygenic risk score, it will be less powerful. Now, we've put a huge amount of effort in genomics to using clever statistical methods and algorithmic methods to reduce that disparity. I mean, at a high level, you want to try and use what we can learn from European ancestry individuals and combine that with the with the smaller amount of data on other ancestries, and you do much better. So, you know, we worked, we're proud of the fact that our prediction scores are more powerful than than those that in the academic world and that other companies have developed. We're also incredibly proud of the fact that our scores in non European ancestry are also more predictive. They're still, in general, there's a gap scores are a bit less predictive in individuals of Asian ancestry and a bit less, even degradation slightly more for people of African ancestry. That's a bad thing. And it needs to change. And it will change through better data and the genetics research community, for example, belatedly, but but to the huge credit, you know, there are massive efforts underway there. The fact that the scores are less predictive doesn't mean they're not useful. And indeed, there are some examples. So so the genic scores matter. But there are other things that differ between ancestries in terms of disease, some diseases are just more common in ancestries and others, some ancestors and others. So for example, in the UK, a polygenic risk score for diabetes, type two diabetes, that kind you get in adulthood, typically, it's a bit less predictive in people of South Asian ancestry than a European ancestry, but the disease is much more common in South Asian individuals. So for a European, someone like me, with European ancestry, if I were in the top few percent polygenic risk, my lifetime risk of diabetes is maybe 40%, without polygenic risk score, because you worked hard on making it as powerful as we can, if I had South Asian ancestry. And I mean, the same top few percent lifetime risk of diseases 75%. Right. So So another thing we put a lot of effort into, which needs care, and I think is people often think the polygenic risk score is the answer. It's not it's a component to the answer. You want to factor in different rates of disease in different groups, as well as the differences in the genetics. And we put a lot of effort into that, for all the obvious reasons, it's so important, actually, the diabetes is at play this gave you is one where, you know, I think that approach will have more impact in the South Asian community in the UK, where diabetes rates are so high than it will in the European ancestry individuals

Patrick Short  28:51  
have different is the strategy for deploying this in somewhere like the US that's very dominated by by private healthcare and insurance versus somewhere like the UK, I have going back and forth while I was thinking of how to ask this because on the one hand, they're they're extremely different systems. But on the other hand, like you were explaining earlier, there's a cardiovascular disease pathway in both systems, and fundamentally, it's additional data point or set of data points to the screening. So are they very different in you have to think about them completely differently? Or, or is it fundamentally the same set of proof points that you need to deliver? And then ultimately, you can roll it out in a very similar fashion?

Peter Donnelly  29:30  
I think there are there are lots of similarities and some different in a system like the UK, for example. It's a single payer system, it's cradle to grave. The National Health Service NHS, as called here is the provider of health care. So if you do prevention better, and stop disease, which will happen in five years time, 10 years time, 20 years time, the system benefits from that, because that person would otherwise be cared for by the NHS, actually for other reasons. I think we're very well placed in the UK the UK has a major effort emphasis on genomics. In healthcare, it's something which is very front of mind for the health system. And for the political leadership, the UK had a very large programme in sequencing for rare diseases and cancers, which was one of the first in the world to do things with that kind of scale. So genomics is is a key part of how the UK thinks about healthcare. The UK even has a genomic national genomics healthcare strategy. And the cabinet minister involved said earlier this week, quite timely for the podcast, but I think he probably had other motivations. He said earlier this year that genomics is the future of post pandemic healthcare. So so that helps in the UK. On the other hand, sometimes government run systems are less agile, and it's less easy to get them to adopt new technologies, the NHS have put huge effort into trying to be nimble with new technologies, that's also helpful. Now, in the US context, as you say, there's a somewhat different system, there are trends in the US which help the trend, the move to value based care is obviously helpful. Some healthcare systems more than others, pride themselves on the sort of long term approach on the prevention that they can offer. So they would be natural places to have early conversations with and indeed, we are doing that and they're very interested for others, you need to be a bit careful. So a natural response from a US insurer is if I spend money now, which will prevent disease in 10 years time, that person may not be with me, and I'll still be covering them. Yeah, yep. So So why do I do it so that that needs a bit more care, those sorts of things, make some things more challenging. But there are other things which are, which are easier. And as I said, the move to value based care helps. I think everyone in the US system thinks prevention is a key part of making healthcare sustainable. And there are particular systems, who pride themselves on prevention, and tend to have longer connections with the with the people in their system, and the ones who are insurance.

Patrick Short  32:01  
I want to touch a little bit on the drug discovery side and therapeutic side of what you all do. In the last couple of minutes we have, because I think it also is a it's a whole podcast of its own and a lot we go in there. But I wonder if you could talk a little bit about how this understanding the human wiring diagram and maybe polygenic scores, in particular play into this side, because people be sick of hearing me talk about PCs, canine and all the learnings from that, that chain. And there's a very clear story there. But the story that polygenic scores tells us very is very much less clear than the monogenic disease story. So what what can we learn from this bag full of pebbles that represents a million tiny cuts that ultimately leads to disease,

Peter Donnelly  32:43  
I think it plays in a number of ways. And actually, if you'll forgive me just for doing that, let me say one more thing on the healthcare front. So we've talked a lot about about these approaches in prevention. And ideally, obviously, you'd prevent disease from happening in the first place. When it does happen when someone gets sick, there will be approaches like this, which will also help us choose the right treatment, right? Whether it's the right drug therapies, or whether for someone who's got prostate cancer, you want to treat them with radiation or surgery, or for someone who has angina, whether you want to treat them with a stent or not, there are situations like that where in the future, these sorts of approaches will help clinicians make choices. So I see in the future genetics helping not just in the prevention piece, but actually once people heal. And then there's also huge potential, and it's what you're just asked about, in genetics help us get helping us get better at drug targets. So the kind of high level and very simple way I think about this is the following. When a drug company is thinking about a new drug, what they're trying to do is to change some aspect of our biology, for example, they're trying to design a molecule, which interacts with a particular protein and stops the protein inhibits the protein from doing what it would normally do, in the hope that that will help someone who has a particular condition or disease. So at the moment, most of those hypotheses come from studies that aren't in humans. One reason we don't know so much about human biology is we can't do experiments on humans, it's a it's a bloody good thing. But we can't do experiments on humans, every other bit of science progresses by doing an experiment, learning something, changing something experiment, learning something new. We can't do that in humans. So it's hard. We use surrogates, we use cellular models or use animal models. So from that, so But suppose I'm a drug company, I've done that sort of work, or I've read academic studies that have done that. And I think if I inhibit this protein, it'll be helpful for say, heart disease. To know whether I'm right, as a drug company, I've got to do two things, but lots of things that two high level things. First one is I have to make a molecule that will get to the right cells, and then latch on and interact with the protein I want to inhibit in the right way, and then stop the protein from doing what it's doing. And on top of all of that, I don't want my molecule to interact with any other proteins because that could have disastrous consequences. So that's hard and takes a lot of time. Suppose I solved that problem. Then the next thing I had to do is to find out whether inhibiting that protein actually is useful to heart disease. How do I do that I do clinical trials, I give it to a small number of individuals who are healthy just to make sure it doesn't have safety consequences, then I do a phase two trial where I give it to a relatively small number of sick people and a small number of Healthy People and see what the impact is. And then I have to do a phase three trial. And for some diseases, heart diseases is one example. Those trials are very large and very expensive. The other thing we know is in spite of the huge amount of effort and really smart people in r&d in the pharma industry, 90% of potential drug targets don't lead to successful drugs, they fail at some point in that process, often the trial process, and many people have argued, and we certainly wouldn't, because we don't understand biology well enough. So here's how genetics and the wiring diagram helps. If you want to know what happens when you inhibit this protein, and whether your molecule will make a difference for heart disease, we can look in our database, because there will be some people who happen to carry genetic variants, which modify that protein, either directly change it, or change things about it. So a common example would be there'll be genetic variants, which mean that in some individuals, the tap that makes the protein is turned down a little bit. So they make a bit less of the protein. So if you stand back a little bit, it's like those people have a weak version of your drug, because they've got a bit little bit less functional protein, they've been on it all their lives, and they've got no idea about it. So it's kind of a perfect clinical trial. And people have called it nature's clinical trial, high level idea is that when we give someone a drug, we're trying to change human biology, genetic variants change human biology, if we can find genetic variants that affect the same bit of biology of your drug, I can do all of that in silico. And I can check it out, you know, we've put a huge amount of effort in to build it bringing together a data resource that links genetic variation with outcomes, I can check in that data, whether the people who's happened to carry genetic variants, which mean that make a bit less of that protein, do they get less heart disease, if they don't, you should be really worried because it suggests that that change you want to make won't have any impact on heart disease. And so we can do it that way around. But you can actually do it the other way around, you can say Abinitio, I want to find find a way that improves heart disease, or diabetes, or arthritis, or lupus, or whatever. And you can say, Let's interrogate the genetics, and see if we can find those little genetic changes that don't make people sick, that they're not a big impact, but they tweaked their biology a little bit. And if we look at enough of people, we can see the consequences. So that's the simple idea behind using genetics to find better drug targets. I mean, it makes sense, I think, and now there's quite a bit of empirical evidence that drug targets that have that kind of genetic basis are much more likely to be successful when they get to clinical trials. So that's one area we can help. But there are others, you can actually use genetics and data analysts in a much more sophisticated way, you can use the genetics as the Pro, which tells you for a given disease, not just which genes, so which proteins might I want to try and change, actually, you can start to piece together the biology, you can say, okay, here are a set of genes which are involved in the same biological pathway, which matter in this particular type of cell for the disease. So the way we think about it is we can use the genetics to identify biological processes, specific bits of biology that happen in particular cell or tissue types that are relevant to disease. So you can say, for a common or rare disease, you can say, I'm going to use the genetics and and a lot of the data that's available these days, functional genomic data, single cell data, and so on to piece together the puzzle, you work out, not just that this gene is relevant, but you can work out that there are a set of genes in this particular pathway in these cells, which matter. And then the next thing we do is we do experiments to verify that, to show that the pretty precise predictions we can make from the data and clever analysis actually holds up that when you do that, you know, not just that I've got to affect this gene. But the genetics can tell you what should change, right? It can tell you what things you could then measure to see whether the molecule you're you're developing actually interacts with the protein, it can tell you what are the right cell types to study, because they're the ones that are actually on the path to disease. So changing a protein in in one type of cell might be fine, but totally irrelevant to disease. So we actually learn the right cell types. So beyond finding the target, there's stuff that's very helpful about the helping find the molecule, the downstream pieces, that readout could be very helpful in a trial. It can be a biomarker that's on the path to disease. And you can learn early in your trial, whether your drug is actually making the biological changes that you want. And then there's a piece on top of that, which is that if you take pieces that you mentioned, PCs K nine, it's a great example, one striking feature of the PCs nine studies. I mean, they showed that if you inhibit PCs K nine, it improves it lowers cholesterol and improves outcomes for heart disease. The groups that did that study in Regeneron. led this also showed that people who have a high polygenic risk score for those people, the drug is more effective. I'd actually other groups say Katherine and his colleagues in Boston, they had earlier looked back over the statin trials and showing you because because One way to go back to our prevention story, it's great to find people who are at high risk of heart disease. If the existing drugs in this case statin have no impact, then you haven't achieved anything. So sake and his colleagues show, firstly, that statins do have the same kind of impact on people with high genetic risk. And indeed, that the data suggests they probably have a higher impact. So for both statins, which is sort of frontline routine approach for heart disease, and PCs, K nine people with high PRs have bigger drugs are more effective for them. Or to put it another way, they have a better gain from the drug. Now, there are reasons to suspect that's more general. So actually, I see a world in which using genetics cleverly to choose people for clinical trials will also mean that the impact of drugs in the trials is larger, that's usually a good thing. I mean, that a drug company might say, well, I don't want to limit the market. But actually, in a world where most drug trials fail, giving yourself the best chance in the first trial is probably helpful. And then you can do a larger trial with a wider population. So I mean, I see and we see and it's central to what we do as a company genetics playing in at lots of places in the drug development pipeline to make a difference.

Patrick Short  41:04  
It's almost endless, isn't it when you build when you build that kind of platform, you everywhere you look, there's some kind of opportunity and and you mentioned, it could limit the population of patients. But I think it could also work in the opposite direction. If you take PCS canine or you know, ApoE e4, Lark to any of these near monogenic or highly penetrant single genes, if you're able to use polygenic scores to identify other people who looked like that, but don't have apo e4. For example, suppose there's a drug that treats Apple for Alzheimer's, you could actually expand the the target as well, because there may not be people who have that specific, that specific snip, but if they have a set of lower impact, snips that converge on the same pathway, you actually you actually could potentially treat those people as well. So I think it has potential to cut the other way from a trial inclusion standpoint.

Peter Donnelly  41:55  
Exactly. So that's for shattering a world in which we're treating pathological biology. Right, rather than disease. Yes, so someone might have this disease, because they've got problems in this bit of some pathway. And we might want to treat that. And I know people who look at at neuro degenerative diseases, for example, they're starting to think in those terms, that actually there might be similar pathways. In some cases of neurodegeneration that are classified as Parkinson's and some that have Alzheimer's, and so on, there'll be a set of, of pathophysiological pathways, and you might want to develop drugs that treat those rather than the disease in general. And so the genetics will help you work out for a given patient, which category they're in.

Patrick Short  42:38  
Very exciting. As a final question here, I'm gonna, I'm gonna ask you to go out on a limb here and make a prediction. I won't hold you to it. But I'm curious on your thoughts, rough order of magnitude, when and I have to be very specific. Otherwise, these predictions are meaningless. But in the UK, every let's say every adult age 40 gets a genetic test and a readout for some some sort of risk of common diseases. If you can make a rough prediction as to what year that would become available as standard in a in the UK, what would you put your finger on, on when, when we'll see that happening?

Peter Donnelly  43:12  
I think it'll start to happen within five years, or about that timescale. And I hope it'll be happening as a consequence of the work we're doing. And the studies we're doing the question of when it will be rolled out to the whole population with the questions when when will it be offered to the whole population? Yes, some people will choose not to go that route. And that's completely fine. That's it's their choice. I would be pretty well, I should caveat this for listeners by saying I'm an optimistic person. So they might want to filter I'm pretty optimistic that it will have happened within 15 years. So probably 1010 years won't be that won't be universal, but 10 to 15 years, I think, I think, you know, I think by then there will be a world where genetic information as a key component, but it'll be all about sophisticated risk prediction. Yes, driven by genetics, and many other things and sophisticated algorithms, I think that'll just be a lot of what's under the hood in healthcare in in 15 or 20 years time,

Patrick Short  44:09  
we're in the things that you and others will be able to do. It's gonna Snowball from that point, right, as this kind of data is collected on Yeah, millions 10s of millions linked with outcome data that all these you know, all these ideas about where in every step of the process we can make an impact will just become easier and easier to do. Right? It's a real Yes. As

Peter Donnelly  44:29  
as, as we start to do this in healthcare systems, the data will be more powerful. The algorithms will get better, we'll be able to develop algorithms for things for which there isn't good data. Now, as you say, it's it's a really virtuous circle, and incredibly exciting and fun entrepreneurs trying to make a difference.

Patrick Short  44:45  
That's right. Yeah. And you know, what you said about being able to make an impact beyond the research. I think it's a really important point. It takes everything from academic research that is very blue skies and unable to see See round corners and make some of these big breakthroughs all the way through to the translational stuff that gets it to patients. So it's it's great that you're, you're dedicating your your time and energy to this really important mission.

Peter Donnelly  45:09  
Thank you very much.

Patrick Short  45:11  
Well, I think that's a great note to end on. Thank you. I really appreciate you taking the time on this Friday afternoon, evening. And yeah, just really appreciate all the great ground recovered. I learned a lot always do whenever I speak to you.

Peter Donnelly  45:22  
It's been a huge pleasure, Patrick. Thank you. Great one. Thanks,

everyone, for listening. As always, please share the episode with a friend if you liked it, and leave review on your favourite podcast player to help other people find us. Thanks a lot, and we'll see you next time.