15
Examination of witnesses
Dr Robert Goldstone and Professor Ewan Birney.
Q11 The Chair: Welcome to our second session this morning. I am very pleased that we have as our two witnesses Professor Ewan Birney, who is the interim director-general of the European Molecular Biology Laboratory and is appearing particularly in his capacity as a non-executive director of Genomics England; and Dr Robert Goldstone, who is the head of genomics at the Francis Crick Institute. By way of opening statements, can you briefly set out your backgrounds and current roles in genomics? We know that the cost of sequencing, and now editing, genetic sequences has fallen rapidly in recent years. Can you set out for us where the science and technology has got to in terms of our understanding of the human genome and what this might mean for patients?
Professor Ewan Birney: Thank you for this invitation. Let me explain the institution I am part of. I am the interim director-general, or head, of the European Molecular Biology Laboratory, which is an intergovernmental organisation based in Heidelberg, Germany; it is headquartered there. It is a part of Europe that the UK did not leave. It co-founded it back in 1974. It has six sites across five countries. One of them is here in the UK: EMBL-EBI. I used to run EMBL-EBI and will go back to running it. EMBL-EBI has the world’s largest and most comprehensive set of databases on biomolecular data. When people use the human genome, the reference human genome comes from EMBL-EBI, and the plant genomes that Baroness Willis mentioned are stored happily in EMBL-EBI.
As well as that role, I am also a researcher. I was one of the people who analysed the draft human genome in the 2000s; I have been with genomics all the way through since then, so I have done research. I have been involved in the delivery of these data resources and data infrastructure. I was on the UK Biobank steering committee. I have helped support a number of countries across Europe in how they think about data, including Denmark and Finland. I am a non-executive director of Genomics England. A final aspect here in which you might be interested is that I am still research-active. I have an active research group, and we are using AI in many different ways. We recently wrote a very successful paper on applying AI to epidemiology—basically, bringing AI to epidemiology and the understanding of risk.
On your questions, genomics is one of the major ways in which we can measure life. Every living organism has DNA, and it is a very useful way of starting to understand life. Of course, there are many other things about living organisms that we would like to know, such as their RNA content and their protein content. Sometimes, we use “multiomic” to mean the collection of technologies, often using DNA read-outs. At the end of the day, we are doing DNA sequencing to measure levels of RNA, protein or metabolite. There are other technologies there. The other big driver for data on living systems is imaging, which has undergone a huge revolution over the past 20 years and is also a big part of what EMBL-EBI delivers.
Dr Robert Goldstone: Thanks very much for the invitation. I am a bit of a minnow compared to the other titans you have invited today. I head Crick’s genomics core facility. We are a technical platform. The researchers at Crick come to us to access our equipment and expertise to turn their biological questions into genomic experiments that can produce the data they go on to analyse in their discoveries. We do not have a research programme. Our team does a lot of method development and technology scoping to make sure that we have the best equipment and the latest techniques available for Crick researchers to move their work forward.
In answer to your question, across the past 10 years or so, genomics has undergone a quiet revolution since the Human Genome Project originally discovered quite how sparse the coding regions of DNA are; in among them was what was thought of as junk DNA. Now we know, from various projects such as the ENCODE project, that almost all of it has a biochemical function. Almost all—90%—of the diseases found in genome-wide association screening come in this dark genome. Uncovering that is moving us closer to understanding the totality of how complex the genome and its regulation are.
That is underpinned by technological leaps now. I cannot even begin to imagine how much it must have cost the Human Genome Project to sequence a single genome. We are getting into—it is not quite there yet—the region of around £100 for a single genome, in terms of the consumables cost. This allows you to sequence at a massive scale, as Sir Mark talked about before.
The Chair: That was a very helpful introduction from both of you.
Q12 Baroness Jones of Whitchurch: I do not know whether you heard all the evidence that Professor Caulfield gave. He has his own expertise, but it would be useful to hear what you think the excitement of the applications might be. Where is it taking us? Professor Caulfield talked about its applications in cancer, but do you go beyond that? Are you already seeing new applications that are exciting you in personalised medicine?
Professor Ewan Birney: One thing to emphasise here is the excitement in a whole range of things, right from basic biology and understanding the fundamentals of life. It is a continuum all the way through clinical research and into clinical practice. When we think about those applications on the clinical practice side, cancer is one area. Rare disease, which Mark also mentioned, has been a tremendous and remarkable genomics success story. It is the right way to diagnose and, where there are good treatments, to assign the treatments for rare disease.
Common disease is an interesting one to crack. There are people doing that both in the UK and around the world. I think that a combination of technologies, one of which will be genomics, will inform the way in which we do common disease, often synthesised by AI. Covid showed us the importance of genomics in infectious disease. That has continued in how we think about TB, HIV and other things. It is about the fast response in a pandemic or an epidemic, where you want to understand where an infectious agent has been and what is going on, as well as in the incredibly practical business of which antimicrobial resistance gene a particular bug has. You can move the point of choice earlier by sequencing the infectious agent, which is exciting. There are quite a few bits of effort along that.
I have previously made an analogy between the uptake of X-rays in clinical medicine and the uptake of genomics. There are some interesting parallels. It is a 30-year story from the discovery of X-rays to their routine use in clinical practice. We have a similar story here for genomics entering clinical practice, and we are 20 years in.
Dr Robert Goldstone: I would add that, in personalised medicine, even if genomics cannot identify the right drug that you might want to use for a particular cancer or disease, it can illustrate what is not the problem and avoid you treating someone with the wrong drug: if they think that it could be an autoimmune disease, they may give a high dose of steroids, but the genomics may say that it is unlikely to be an autoimmune disease. You can avoid the harm that is caused by drugs that are misprescribed.
Baroness Jones of Whitchurch: Are you beginning to see some red flags in where it is not applicable already? At the moment, we are all talking in a positive way, but there must be limitations on the applications of this as a diagnosis tool.
Professor Ewan Birney: X-rays and imaging are a very good way of anchoring this. There are diseases where X-rays are totally transformative to the diagnosis and what will happen next. There are other diseases where it is not useful at all. I think that genomics will go into the same zone. In general, most diseases will have some information from the germline genome, the cancer genome or the infectious agent because a disease usually touches it. Sometimes, it will be very limited; it will be just to rule things out, as Robert mentioned, rather than to rule things in. I really doubt that it will be completely uninformative in future.
Dr Robert Goldstone: Sir Mark was talking about a patient where they sequenced the genome and found many thousands of mutations. They were able, in that case, to limit it to one mutation that was the cause of the disease because of an absence in the parents’ genomes. That is not always the case. Sometimes, there are variants of unknown significance. In a sense, you can often drown in data and not know what to do with it. That is where the technology is slightly limited right now, in that we do not know what all of these mutations are doing.
Q13 Lord Duncan of Springbank: Picking up on the drowning in data point, sequencing has become quite mature as a way of understanding, but interpreting the variants and the elements must be quite a challenge in terms of having the sheer ability to have people who can do that. Is that the bottleneck that will cause the problems in taking it to the next stage?
Professor Ewan Birney: It is computational biology. This is my field, and I am not going to be out of a job. It has been a challenge. It feels like climbing the Himalayas or something: you make a lot of progress but the mountains still seem just as high. You have to turn around and look back to see where you have come from and how high you have gone before you turn around, look forward and scale a bit more of the mountain. We know so much more about the human genome now than we did in 2000, yet we still have many variants of unknown significance—that is, regions in the genome that are clearly involved with the disease but we do not know why. There are lots of unknowns. It is true that AI gives us a new tool and a new view on that—a lot of progress from AI has stimulated it—but it will not be solved tomorrow. This is a multiple-decade problem, perhaps an evergreen problem. You are talking about understanding life, and understanding humans in their entirety, when you talk about this challenge.
Lord Duncan of Springbank: Is the NHS putting enough money into that bit of the challenge? Can the personnel and staff tackle the mountain?
Professor Ewan Birney: My expertise is focused on basic biology and the way in which we make a bridge into clinical research and clinical practice. It is tough to ask somebody like me whether we are science-limited, sample-limited or money-limited. We are definitely money-limited in this game. In terms of how you tension that against other things, I am very glad that I do not have to make those decisions.
The Chair: It is the big question.
Q14 Lord Patel: Thank you for coming today. There is a lot of excitement about the use of AI in the development of AlphaGenome. I would like to hear your views on where AlphaGenome, now that we know about the folding proteins in the human genome, is likely to go. Where is the next development and what will it require? What will be required to be invested in it for its potential to be delivered?
Professor Ewan Birney: AI has fundamentally changed how we think about a number of biological problems. AlphaFold, which was created by the team put together by Demis Hassabis and John Jumper, solved a 1970s problem and rightly got the Nobel prize for that. AlphaGenome, which comes from Google DeepMind—Žiga Avsec is the lead author—is a very good piece of AI, but I prefer the models coming from some researchers in Stanford. They are given slightly less sexy names, such as ChromBPNet, but they are just as good, in my view. I am sure that Žiga would say the same; I know Žiga well. We use AI to tackle some novel problems, and the future there is very bright.
What do you need for good AI? You need good data. If you do not have the data, you will not be able to produce any sensible piece of AI. You need good people—that is, people who are skilled in the art of creating these different methods and developing and testing them. Testing them is half the problem over and beyond developing them. You also need enough compute. Some of these methods are quite compute-hungry. An interesting thing is that you need the right questions that you want to answer; your choice of which things you study is important.
You need those four things, a lot of which involve skilled people: people who can do the AI, manage the data, run the machines and formulate the right questions. Many places in the world do this. It is a big part of our future in science, but we should be optimistic about it. There will be other problems that we can solve in this space, for sure.
Lord Patel: Where are we in the UK on those four things?
Professor Ewan Birney: Europe as a whole is in a strong place across AI, particularly on data. As I mentioned, at EMBL-EBI, we have the world’s most comprehensive set of biomedical data resources—it is here in the UK and is part of a European federation via EMBL—and there are a lot of very skilled people. There are lots of clever Americans and clever Chinese people. In the basic research space, we share everything every night. We synchronise datasets—not the human population datasets but the reference datasets, the plant and animal datasets, the protein-folding datasets and such things. That community works well together globally—China included—in delivering this. I do not use a strong competitive lens between countries here. I take a much stronger view: what are the problems that we can solve and who is best placed to solve them?
Lord Patel: Dr Goldstone, you work at Crick. Where are we in using AI and machine learning, focusing on the use of AI to develop healthcare benefits? For instance, where will the folding protein story go next?
Dr Robert Goldstone: In contrast to Professor Birney, I am more of a wet lab biologist, so I come from a different end of the spectrum. In our hands, AI has transformed basic research over a number of years. Even going back 10 or 15 years, we were using AI to predict functional proteins’ gene function; these days, we are using AI to call bases when we are trying to sequence DNA and stuff like that. So we definitely come at this from much more of a basic research aspect.
On where things are going for medicine and the NHS, I mentioned these variants of unknown significance. AlphaGenome allows you to identify, if you change this base of DNA, the likely effect of that change on an area of the genome around the base in a megabase range. How might it change the expression of genes around it? How will it change the accessibility of that chromosome and DNA to other proteins that get in and change the expression of different genes? It is a powerful tool to simulate what a variant might do in terms of causing disease. As Sir Mark said, if you can say that a disease looks like it has some metabolic underpinning and see that this variant affects the expression of metabolic genes around it, you can more easily determine—or have a likelihood of saying—that this is now the causative variant causing this disease, in among the 6,000 or so variants that we were discussing before. That is where it is going right now.
Lord Patel: I am crystal ball gazing. How will the information that you get from AI about the genome and the biological information it produces, such as on alpha 4, help us develop medicines, diagnostics or mutations?
Professor Ewan Birney: It would be useful here to talk briefly about a piece of research that came from my group and a colleague in Germany, Moritz Gerstung.
Lord Patel: Was that the prompt?
Professor Ewan Birney: We repositioned the same technology that underlies large language models but, rather than trying to predict words so that a chatbot talks back to you, we try to predict the next health event that someone would have in a hospital. We prompt it with a series of events over someone’s life, which are recorded in their healthcare records, then the task we set the AI algorithm to solve is, “What will be the next event for this individual in a hospital?”. We trained with the UK Biobank, which is a wonderful dataset to use, with half a million Brits. We tested on a held-out dataset from the UK Biobank, but our most impressive test was when we went to Denmark and tested it on 1.9 million Danes. Our training system had never seen any Danish data. We could predict reasonably accurately, at least in the short term, the total disease burden of over 1,263 diseases for individuals. We also do that using current clinical epidemiological methods.
Interestingly, in the first version, we did not put in any imaging, blood biochemistry or genome. We are now putting those in. This is an example of the capabilities that AI gives us. Previously, it was considered impossible to consider all diseases at the same time. It is not; it is perfectly feasible. Given the right datasets, which we have in the UK and in Denmark—probably in Finland, too; my colleagues in Finland are doing this—we can train these models, test them and have confidence that they generalise. We can have confidence that the UK model works in Denmark or a Finnish-trained model works in the UK, which gives us confidence that we are learning about how humans interact with healthcare systems and biology.
Lord Patel: Is it not correct that, to learn the progression of a disease, you will need not only information that the genome might give you but epigenetic data?
Professor Ewan Birney: Yes. The model that we have does not have any genomics or epigenetics in it. It is just your life history. In some sense, it is just epidemiology—it is a straight-up epidemiology model—but we are excited. One of the benefits of using AI is that we can fuse different data sources very easily. Fusing imaging with healthcare records, epigenetics or blood biochemistry is all feasible; those are active research topics.
Dr Robert Goldstone: The only thing I would add is that, in talking about how important it is to have good-quality data, the data is absolutely essential. However, as important is for that data to have good metadata around it. Unless you have all the conditions under which the data was generated and the instruments with which it was generated, you do not want to be training your AI models on noise in that data. You want to know exactly what is going into it. EMBL was ahead of the curve when it comes to strategy around this, and institutions in the UK are playing catch-up.
Q15 Lord Ranger of Northwood: Thank you both for coming in. It is interesting that you are talking about the models, the data and geographies—Denmark and the UK—because there is a sovereignty issue with data. What does each country have that you can work with? You have said that you need good data and compute. In the UK, what should be our priorities to ensure that we have the right data infrastructure to advance the ambitions that we are now seeing at scale? I spent a lot of my career working in the technology sector with managed service providers; I worked with a European provider called Atos. Is there a model that you are looking at which can help provide, in partnership, the infrastructure and data requirements that you might need going forward, considering the size and scale of what is required?
Professor Ewan Birney: Gosh—that question is quite big. It is very important that the UK continues to play its role in supporting this basic science, research and data storage. Some of that is the data generation, which happens in institutes such as Crick, but another aspect of it is supporting places such as EMBL-EBI, which is the organisation that runs the databases here in the UK. For example, we store the reference human genome and update it every three months, with latest knowledge being generated from that. There is a substantial disc part there; the UK Government have been very generous in how they have supported that, to their credit.
You mentioned data sovereignty and other aspects. Certainly, when you get to humans—tracking patients and citizens—nearly all countries take a view that population-scale datasets must be kept within the country. That is incredibly valid; it is the consensus view on how we handle this in different countries.
The UK has opportunities here. It is a pretty large population, and NHS hospital records were made electronic in 1992. In Finland, they did it in 1982; in Denmark, it was 1977; and Iceland is somewhere in the 1980s, but it is very small. Those are the only countries that go earlier than that, so the UK is certainly one of the largest countries that has such a long electronic healthcare record period—at least for hospital admissions. Also, GP data has recently been changed; something important, possibly legislation, has changed around GP data in the UK.
That has to be operationalised, though. Saying that those datasets exist is not the same as making them accessible. The interesting new aspect here is the HDRS—the Health Data Research Service—which is a UK Government-announced company with the involvement of the Wellcome Trust. I am very excited about the possibilities there; it is very positive. That would complement the world-leading UK Biobank and Our Future Health, which is the next cohort coming along. There is lots of positivity but many a mistake could be made in operationalising these assets.
Lord Ranger of Northwood: Do you see anything emerging in terms of partnership models?
Professor Ewan Birney: That is not my end of the business, as it were. I am much more focused on the basic research in this clinical space; that is the place where I have expertise.
Q16 Baroness Willis of Summertown: You have mentioned Finland and Denmark, but how do we compare with other countries? Are we still a leader in this area of data? Where are our gaps? Where would you like to see us in five years’ time in terms of data management?
Professor Ewan Birney: I am paid for by the taxpayers of 30 different countries so, when I use the word “we”, there is very much a European “we” about it. I would like to answer that question from a European perspective. Europe has real strengths in data—it truly does. This is a definition of Europe that includes the UK, which is a big and important player in the data space for many different reasons. I love my Brits and using populations of Brits, Danes and Finns. As a researcher, those are my three datasets: Brits, Danes and Finns.
Baroness Willis of Summertown: Following on from that, thinking about barriers, you will have heard the question I asked before about the Amazon cloud. Do you see that as a threat going forward, in terms of the increasing cloud costs and the sovereignty around who owns or manages that data? Someone could flip the switch. There are so many issues there. Do you have banks of data servers?
Professor Ewan Birney: At EMBL-EBI, it is a small-scale hyperscaler. We run about 0.3 exabytes of disc, so we achieve scaling at a very low cost. It has been hard for us to get similar costs in the cloud; we have looked. I have nothing against it, and we do use it for many things. There must be commercial, cost-driven views on this.
You mentioned security, sovereignty and other things. It is a two-edged sword. There is no doubt that big cloud providers have a level of computational security that is hard for anybody else to match; they have more muscle and depth. However, there are these appropriate concerns around sovereignty. Sovereignty interacts with where things are physically, importantly. Most cloud providers will give you UK-sovereign clouds and will give you legal security. You can then think very hard about how, if they are a US company, the US Government would work with them. Eventually, you have to take a view on this, similar to the view you take on your mobile phone, Microsoft Word or the Macintosh you have.
I would not want to blow that problem out of the water. We will not own our own Silicon-up stack, as it were, so we have to make decisions on this. One can handle security on the cloud with good contracting appropriately, but you do need good contracting. There is no doubt that there is a commercial side to this.
Baroness Willis of Summertown: My thought is that, yes, you can have this data on a UK cloud, but—
Professor Ewan Birney: Via UK Amazon or UK Google.
Baroness Willis of Summertown: But the person who manages the overall infrastructure is not in the UK, ultimately.
Professor Ewan Birney: As I said, this is the legal contracting that you have to go for. You could be super paranoid and say no—
Baroness Willis of Summertown: I think that we have reason to be so right now.
Professor Ewan Birney: There are appropriate decisions to make. However, one does have to live in the world where you just cannot buy all the kit you want to buy and run it yourself from the UK.
Baroness Willis of Summertown: You have taken the decision not to be on the cloud.
Professor Ewan Birney: Only because we cannot match the costs.
Baroness Willis of Summertown: So it is an economic decision.
Professor Ewan Birney: Yes. The data we store is open to everyone anyway because it is like the public human genome and all these public datasets, such as for your lovely plants. We do not have the security concern. Genomics England, for example, does have a security concern, which is in a completely different place from EMBL-EBI.
Q17 Lord Drayson: Professor Birney, it is excellent that EMBL-EBI is in the UK, despite the madness of Brexit still prospering. You have an open source business model, is that correct? In effect, you make your assets available to whoever wants to use them for research purposes with no charge.
Professor Ewan Birney: That is correct. It is not a business model; it is more that we are a public good supported by 30 different countries for public synergy. In that open source, a very important piece of reciprocity is that the American and Japanese systems are also open. We all swap our data every night. So there is a piece of reciprocity there, across global reciprocity, in the way we do data sharing, as we have done since the Royal Society was founded here in 1660 and published scientific studies. It is that ethos of open science that drives it as a public good.
I emphasise that we are very open, collaborative and porous with business. We want to make businesses stand on that same dataset. We have a number of public-private partnerships where businesses deliberately put a lot of cold, hard cash in so that this part of the public domain becomes robust enough for them to stand on.
Lord Drayson: In your role as a non-executive director of Genomics England, you heard my question to Sir Mark earlier.
Professor Ewan Birney: I did.
Lord Drayson: Have you smelt the coffee?
Professor Ewan Birney: My role on the board is definitely on the basic research to clinical side. There is a lot of emphasis on the commercial side, appropriately. I am also a shareholder in and long-time consultant for Oxford Nanopore; that conflict of interest sometimes takes me out of the room in Genomics England discussions. I was pleased to see the different deals that Genomics England has struck with Oxford Nanopore on different collaborations. On the businesses that have come to Genomics England—you would need their commercial director; I cannot name them off the top of my head, I am afraid—there definitely have been success stories.
Your point about striking the right balance is, I think, valid. It is complicated. I frame this as, “It is really important to support businesses”. How does the UK taxpayer gain benefit from this? It is by having businesses move here and stay here, and by having people from businesses that want this working here, rather than cash being handed to Genomics England. That will be less revenue-generating, I suspect. Commercial companies expect to pay—they often want to pay if they see value in the datasets—so it is important to get that business side of Genomics England right, as well as all these other sides. There is definitely an art to how one does this.
Q18 Lord Winston: I am glad that you brought up epigenetics; I thought that we might go right through this without ever mentioning the action of the environment on genes, for example. In particular, the field that I think is the most interesting is developmental medicine. Much of that work will take much longer to unravel. The issue then, of course, is cohort studies, in which you have great expertise. Do you think that we are doing enough to look at those sorts of issues with cohorts?
Professor Ewan Birney: The environment or, in particular, the environment in early development—both. First, I am a geneticist and, in some sense, a genomicist, but I get slightly frustrated by my colleagues who put their blinkers on and refuse to look at the environment because it is too complicated. They stay focused on the genomics. We have to take a holistic view of all the inputs that come into what makes an individual an individual, and lots of things that are non-genetic change the life-course of an individual. Some of my research is in precisely this area, by the way.
A good opportunity here with AI, particularly in the work I mentioned, is that we can finally fuse very disparate data sources together. We can fuse environmental, basic epidemiology, blood biochemistry and genomics all into one model. An interesting problem now is what bits of the environment we should measure and how we should measure them. Humans are very social animals, so, when we talk about the environment, do not forget that this involves all of the social aspects: how you get to healthcare, how you access it, how your family handles different healthcare events—those sorts of things. That is a very important part of the environment for a human, especially as they are growing up.
It is quite a big box to open, if you know what I mean, but open it, nevertheless, we should do. It basically involves, first and foremost, capturing the data on these large-scale cohort studies of all the above. You do not get any good AI without good data; that is where you have to go. I advocate for good, sensible, cost-effective ways of adding environmental information into these large-scale cohort studies, reminding ourselves that the environment here includes the social environment, so you get into sociology and some pretty straightforward socioeconomic aspects of life.
Q19 Lord Winston: Dr Goldstone, the Chair mentioned something important in his last question: having clinical academics. It is an interesting issue because you are at Crick. I wonder whether you feel that there is enough going on between medical schools and Crick. Can you do more about that? To what extent are you concerned about the fact that so many basic scientists—including some of the people I work with in my work at Imperial College—are not interested in the idea of investment? They do not do science for money, and the idea of economic growth is almost beyond their interest. Indeed, that was almost a problem with UKRI; it was launched in a slightly different way when it was produced, which was not always a great stimulus to the idea of trying to promote what we have been hearing about from Lord Drayson.
Dr Robert Goldstone: There are two questions there. The first is about having a workforce that is multi-skilled between clinical understanding, biological understanding and computational biology understanding. As I mentioned before, I am a wet lab biologist. I still break into a sweat if I see a statistical model; I am sure that some computational biologists feel the same if they see a diagram of a cell. It is a rare breed of person who can master both or all of those arts in one go, so having closer collaborations between these people to bring out the best in them is important. Places such as Crick do that; it has been a real success story. Our core facility works with a number of clinical researchers at Crick on all sorts of different research topics. They bring such a unique perspective to the work, then we bring our side of things and the bio practitioners bring their side of things. From that union, you get some interesting science.
That segues somewhat into the second question, which is about commercialisation from research. You are right: undeniably, a lot of researchers do science for the sake of science. They want to discover things for the sake of knowledge rather than because they want to create a drug or they want to commercialise it in one form or another. I wonder sometimes whether that is not a bad thing. The classical model of treatments for diseases that people have to take for a lifetime is failing now, and we are moving further is preventive medicine and preventing you becoming ill in the first place.
Those discoveries are created by that foundational research, rather than someone having a commercial mindset. I would not want to devalue the fact that fundamental research is done for the sake of discovery, because that is moving us more into this world where preventing disease is more important than treating it.
Q20 Lord Patel: You have referred to several bodies that would be involved—health research, data centres, Genomics England, genomics services, et cetera—to try to deliver both research and healthcare. The 100,000 Genomes Project, big as it is, is small as compared to, as Professor Sir Mark Caulfield said, 67 million people’s health data for the purpose of research. Who in the UK would be able to run such a huge data model?
Professor Ewan Birney: Due to the fact that the 55 million English dataset was made accessible for responsible use in Covid research during the pandemic, through the BHF facility, we can definitely use those datasets and get insights from them. Researchers have already shown that it is feasible to get lots of useful information out of that. In a recent piece of work we have been doing, my estimate is that we need 10 million to 100 million-sized cohorts to tackle Lord Winston’s gene-environment interaction studies. Those are the data sizes we need. We should not be scared of these data sizes; they are computationally feasible to handle. Doing that does not scare me.
The challenge is operationalising this, as I mentioned. During Covid, emergency data access routes were set up, and there was a lot of good will in the system that allowed epigenetic engineering and other things, but it was a one-off and was not a completely stable, solid system.
I view the HDRS—this goes to Cathie Sudlow’s report—as making that a more routine part of how we think about this fusion of practising healthcare; running healthcare, which is namely what we would call the NHS; and clinical-level research into these things. We no longer think of discrete cohorts; we think of the entire population. Cathie Sudlow’s report is a good report here, and that BHF-based Covid analysis is proof that we can actually do this. There is no technical problem in being able to do this.
Lord Patel: This is a challenging question. In answering Lord Winston’s question, you referred to the importance of epigenetic environmental information and the difficulty of capturing that. You suggested that it is okay to rely on genomic data, imaging data and metabolic data from blood samples or whatever—
Professor Ewan Birney: And lots of environmental data.
Lord Patel: Right. So why do we not use AI to collect this environmental data?
Professor Ewan Birney: AI by itself will probably hallucinate very real data but it will not be very useful. For AI to be—
Lord Patel: It is not afraid of big data, so you can remove the guards.
Professor Ewan Birney: AI is already very useful in this metadata upgrading, where you take dirtier data and clean it through a variety of processes, but AI by itself cannot create observations about human beings. Someone has actually got to do the observation of human beings and correlate, “This individual was there with this environment doing this with this aspect”. That fundamental piece of measurement cannot be substituted with an AI model.
The Chair: I think that we should move on, Lord Patel. Have you finished, or do you have more questions?
Lord Patel: I am finished.
Q21 Lord Booth: Thank you, gentlemen, for coming in today and for your evidence. At the end of the day, we will produce a report. It would be helpful for us if you could tell us what your specific research priorities would be for the Government to support the development of personalised medicine. That would really help us when we come to produce the report.
Professor Ewan Birney: It is quite an open question. I can give a narrow answer and a broad one. The narrow answer is that it is important that the UK Government continue to support this basic-to-clinical infrastructure aspect. The human genome is used for both basic research and clinical delivery, and there is no difference in understanding humans. You need to understand how human biology works from DNA up for both tasks. I am glad that we have not separated those tasks and that the research side and the healthcare side continue to support that effort jointly. I would very much argue for that to continue.
In addition—this is with my much broader hat on—the HDRS is an important initiative that has recently started. I worry above all that we have put too many expectations on it for its first couple of years. We need to give that organisation time and space to start up, find its feet and do the right things, without insane expectations all around it in that very early place. We therefore need to optimise the first five to 10 years, not the first two or three years, in that space.
Those are my two concrete proposals, the first one being somewhat more self-serving.
Q22 Lord Willis of Knaresborough: I want to return to some basic questions, which I tend to do in this committee. Do you feel that, in terms of our healthcare system, particularly in the NHS, we have people who have enough qualities to do the research that enables the work we are talking about to be done? I do not believe that that is the case. We have a real problem in trying to get people working in the healthcare system, choosing to continue to do research as part and parcel of their work rather than simply dedicating themselves to working in hospitals or clinics.
Professor Ewan Birney: I am not the right person to answer that question; it would be better answered by someone who really lives inside the NHS well. I have two comments to make. As I work with researchers around Europe in different clinical settings, it is as enjoyable and invigorating working with Brits as it is with Finns, Danes, the French or Germans. There is no obvious lack of quality here. That is one thing.
The second thing, which is an interesting insight I have realised, is that we are blessed by the fact that the international language of science—English—is the same as the language we use to interact with our population. So we do not have a major barrier there. If you go to other settings in other countries, there is an interesting problem that we do not face here in the UK in the same way: the people doing healthcare practice often do not have English that is as good as that of their clinical or basic research colleagues. That is an observation about how lucky we in the UK are that the international language of science is English.
The Chair: We are coming to the end of our session. The last question is from Lord Duncan.
Q23 Lord Duncan of Springbank: Obviously, this will result in a report, primarily to government. I would like to try to understand what you think we should be saying to government right now. What should those priorities look like? I am struck by the conversations we have had about a number of the problems you are experiencing, so what would you say to government about sorting it out?
Professor Ewan Birney: I repeat the importance of basic research fused to clinic, and that continuum. There is, of course, a natural tendency to try to draw clear boundaries in that space, because one has to assign money to departments and budgets, but there is no clear boundary between basic research through to not simply clinical research but clinical practice. The human genome is a good exemplar of that aspect. Making sure that the UK Government continue to fund basic research through to clinical practice and clinical research is incredibly important.
Self-servingly, the UK is wise in its investment in places such as EMBL; it is great that it is supporting EMBL-EBI here with the capital. I would like that to happen again in future. We have another big deep breath coming in 2030 so I am focused a bit on that.
Taking that hat off and putting on a much broader UK hat, I want to double down on the HDRS as the new kid on the block in terms of what is going on in the UK. My advice is to give it space to find its feet in years 1, 2 and 3, and not to have too many expectations. Something that worries me there is not the strategy choice or the people, which are good, but the expectations around the organisation already being completely unrealistic, from my perspective. Everybody needs to give that organisation time to boot up, find its feet, get the right people and get its head around what it is doing. Then you definitely need to invite Mel Ivarsson and Nicola Blackwood here to give evidence; they would be the right people.
The Chair: Dr Goldstone, do you want to add anything?
Dr Robert Goldstone: The only thing I would add to that Sir Mark was talking about multiomic analysis and you were talking about having different observations. Providing the tools to make those observations and measure those multiomic features requires the technology and the methods to do it. One thing we struggle with in our organisation is the funding to develop those new tools and methods. If I could say that there was a priority from my side, it would be UKRI having funding sources available for that method development. It is hard to get hold of that in the UK. You can get money for equipment and new instrumentation—that is fantastic when we need that—but you also need money to develop the next generation of technologies, in addition to buying the existing ones.
The Chair: Gentlemen, thank you very much for a very informative session. We are very grateful to both of you for coming. That concludes this session.
