In this episode of FoGCast, we sit down with Adam Cribbs, Associate Professor at the University of Oxford and co-founder of Entelo Bio, to explore where biology, machine learning, and medicine truly intersect. From retraining across disciplines to building discovery engines for drug development, Adam shares a refreshingly pragmatic view on AI as a tool, not a replacement, and what it really takes to drive impact in modern science.
Lyndsey: Hi everyone, and welcome back to FoGCast. Today, Manny and myself are sitting down with Adam Cribbs, Assistant Professor at Oxford University who leads a team who are embedded in the latest tech, integrating functional genomics with advanced machine learning approaches and other tech like single cell to understand disease biology. Adam, thank you so much for joining us today. How are you
Adam: Good, thank you.
Lyndsey: Tell us a little bit about yourself. How did you get to where you are today?
Adam: As you rightly mentioned, I’m an Associate Professor in Computational Biology at the University of Oxford. I’m actually part time. I’ve also got another role where I’m CTO of a spin out that I co-founded in January 2023, and the reason why I mentioned that is ultimately because I think that defines my journey through the whole of my career so far. I’ve actually been in Oxford for 13 and a half, nearly 14 years now, and I started my career at the Kennedy Institute of Rheumatology. It was originally in Imperial, but transitioned to Oxford halfway through to my PhD, or actually towards the end. And the reason why I decided to embark on a PhD at the Kennedy was because I was really interested in translational medicine. So, how do you take really basic scientific ideas and actually translate them all the way to the clinic? And the Kennedy Institute is renowned for development of anti-TNF therapy, and it was a fantastic environment to ultimately see how you could take that really, really basic understanding like, ‘hey, TNF is important within metal arthritis’, and actually develop a medicine around that basic core idea. And being in that environment, you could really get a sense of how it’s not just about taking those core ideas, it’s actually building teams around that idea and having multiple different viewpoints on a particular area, from not just the clinic, but also from the basic science as well. I did my PhD in molecular immunology, and then slowly transitioned towards computational biology, because after I undertook my PhD, I realised that computational biology was going to be really critical for the next generation of discoveries, and that’s how I transitioned towards computational biology.
Manuel: What years are we talking about?
Adam: I completed my PhD in 2013 and then I did a year of post doc, and I was very, very lucky to have been accepted on an MLC fellowship to completely retrain. I did have a computational background before I got into science, but it was a three year fellowship programme to learn mathematics, statistics, computational biology, and then combine that with the wet lab skills that I’d earlier developed and generated during [my time at] the Kennedy Institute After that fellowship, I got recruited to the place where I am now in Oxford, at the Botnar Research Centre, where I’ve been slowly building my academic career and combining those different types of skills, because I really believe that discoveries are made at the intersection between different disciplines, and I’ve always wanted to not pigeonhole myself within one particular area. I love coming to work and learning new things, learning new different types of science. And so, I’m not a chemist by training, but a lot of my work intersects computation, chemistry, biology and a bit of physics.
Manuel: You went from a computer science background into immunology. Is that what you’re saying?
Adam: No, I had a self-taught interest in cryptography and computation. I actually learned coding from a book in a library before my parents could afford a computer, and then we bought one from the middle aisle at Aldi, believe it or not, it was a Medion 365 computer. I ended up creating a lot of programmes that annoyed a lot of the neighbours, because it was the dial up internet phase where you could generate code and ring neighbours etc. I learned a lot of the computational skills initially and the coding through my mischievousness at 14, 15, 16.
Manuel: Now that we are entering into AI, people are saying, you shouldn’t do computer science because all the code is going to be generated by AI. Do you agree with that statement?
Adam: I don’t. Actually, I think there’s an overblown sense that software engineers are going to get replaced. But actually, I don’t think that. Scientists aren’t going to be replaced. I think what’s going to happen, is AI is a tool, and people who use AI will replace those who don’t. So, that’s how I see and envisage it. There’s always got to be, at some level, a human in the loop. And the way we use machine learning and AI is to speed things up or compress a particular type of problem. So, you’re not really focusing on the very low level. You’re able to make much higher level decisions. You can change a project in a certain direction without having to review reams and reams and reams of data. You can really use it to compress the information you’re looking at. Because I don’t know about you, but I struggle to see in more than three dimensions.
Manuel: Would you say, based on that, from your perspective, what you’re going to have is a tool that helps you accelerate and test your hypothesis in a much faster way? Would you say that’s the implication?
Adam: That’s exactly what we’re using it for at the moment. I mean, I’ve written a lot of agentic tools to automatically write pipelines. And the reason I did that wasn’t because it’s cool and trendy and agentic systems are the vogue thing. It’s because I just hate writing pipelines. I just find it really boring. Now, I used to find it really exciting when I first moved into computational biology, learning new, different ways of thinking about programming, it was like a game to me, but now I’ve done it for 14 years, it’s time to automate it. And so I’m always thinking about automation. What things can we automate? And what things shouldn’t you automate? There’s always got to be a balance. You can over automate a process, and what that ends up doing is, you’re not really thinking about the problem, you’re getting the AI to think about the biological problems and the biological hypotheses. But actually, we all know in the current state, tools will get better, and things hallucinate. You know, ChatGPT, it’ll just give you a whole load of random references that don’t exist. So, you know, you always need to learn and review the literature yourself.
Manuel: I said that I was here to learn from you. So, what books are you reading right now, about agentic?
Adam: I read a lot more about business philosophy, to be honest. Rather than hardcore books on LLMs, it’s more about the philosophy. One of my favourite books is David Graebers, and this is probably where you might have to edit this out, ‘Bulls**t Jobs’.
Manuel: No, this bit is going to be very important. I need to know those titles.
Adam: David Graeber, unfortunately, he passed away. He was a professor at LSE, and was really studying how companies and systems in society generate these jobs that have no meaning or no real value. There’s a whole philosophy of why these occur. And you know, and I can see firsthand this occurring at the university. To be honest, it’s probably one of the frustrations I have, working in the university system, that there are too many different routes to just getting an easy problem solved, and all you need is just, hey, press the button and let me do it. [Instead, it’s] ‘Oh, no, I’ve got to contact this person, this person, this person.’
Manuel: I’m sure you use those frustrating stimuli to foster change. It seems to me that you like to be a little bit ahead of the curve, particularly now on the agentic system. I mean, I’m sorry to say, you said the buzzword. I didn’t say it, you brought it up. So, I’m going to pull that thread, if you don’t mind. How do you make sure that you are ahead of the curve?
Adam: It’s by not believing the hype. It’s by implementing things very quickly. I mean, you can see the evolution of machine learning over the years, where you have upticks, trendy upticks. It was the perceptron back in the 50s and 60s, that then died after Minsky came out and said, hey, actually, it doesn’t solve an XOR problem. And then the re-emergence after the work of Hinton in neural networks. All of my students always wanted to solve what could have been solved by a gradient boosting algorithm using neural networks. So, I would always let them entertain themselves, but I’d always say, hey, also implement this. It’s a much simpler method, and actually, I bet it will outperform it, and often it did. Sometimes I’m wrong, though, and I like being proved wrong. I don’t have an issue if a student comes to me and says, ‘hey, that was a whole lot of rubbish, actually, this solved it way better.’ And then you put your hands up. But at least I’ve learned from that. And it’s about trying things. But I find that the hardest problem in science is actually giving up on an idea. And I really struggled with this for many, many years, but when I really thought about it I thought, well, fail early, fail spectacularly. I don’t know where I got that. It’s a philosophy from somewhere that I read, and it’s true, because you don’t want to fail after you’ve invested so much time in a particular idea that goes nowhere. So, implement a minimum viable product. Test it, benchmark it, if it works, then really spend a lot of time iterating over it. If it doesn’t, then just give up on it, because you’re just going to waste your time where there’s a plethora of potential discoveries that you might make if you drop that idea. Unfortunately, the most precious thing in your life is time, and you’re not going to get that time back, so spend it wisely.
Manuel: I don’t think you’ve answered my question, how do you stay ahead of the curve?
Adam: Staying ahead of the curve, it’s hard. It’s really hard. And it’s about reading the current state, thinking about what the current state might be in the future and trying to iterate towards that. The problem is, no one can predict the future. So often, you might be wrong, and I’ve been wrong many times. But ultimately, moving towards a particular path, as long as you’re moving forward. [In case] that path is no longer viable, always have multiple paths that you’re also working on at the same time. My students will often get a number of projects, usually two or three. One that’s going to work, and they’re going to get that PhD. It’s not necessarily going to be the most spectacular. Second and third projects are going to be much riskier. We don’t know whether it’s going to work. Actually, the path forward might lead to a new discovery or a new avenue, but it also might lead to nowhere, and the least risky project is usually the one that is embedded in the current understanding.
Manuel: Will funders fund [risky ideas]?
Adam: No, no, no, they don’t. Funders will fund things that are typically boring. I’ve now got a really love-hate relationship with peer review. Actually, one of the reasons why I ended up setting up a spin out company is because academically, I couldn’t get my what I classified as my bold, crazy ideas [funded], but actually they’re much higher risk, much higher reward. So, I set up a company to take all that underlying discovery, turn it into something that will be a discovery engine for developing new medicines in the future. And most of the projects that get funded aren’t necessarily the riskier ones, and it’s down to peer review. You ultimately get a score back from a peer, from a reviewer, and they give you a five, and the rest might be eights or nines. That’s instantly killed your grant. And so for me, unfortunately, in academia, you often have to play the game. You might have one or two small, risky ideas in that grant, but the majority of it isn’t necessarily going to be truly risky, because the system doesn’t allow for it. I do like new initiatives like the ARIA programmes that are coming on board, where actually there are funds for more high risk projects, but the majority of funding is typically not that.
Manuel: What’s the name of your company?
Adam: Entelo Bio.
Manuel: What does it do? Do you have customers?
Adam: Yeah. We are focused on using long read sequencing, generating a multimodal data stack. For me, it’s all about data. If you are working on complex musculoskeletal conditions, and if you work on complex diseases, you need an equally complex data set to really understand what is happening on the biological level, but also at the cellular level. And that is generating different types of omics approaches and then using some of our machine learning to then understand and compress that information down. You can then ultimately understand biological pathways in a lot more detail, or actually identify targets by strategically looking across that high dimensional data and working out which targets might be important for therapeutic development.
Manuel: Are you training your own model, or are you building on top of existing models?
Adam: A bit of both. We treat machine learning like a tool. We are a biology first company – biology first, AI everything. So, anything that is automated, we will do to a point, but for machine learning, it’s really working out what your biological question or problem is, and depending upon the question, we then use different types of machine learning tools around that. So, a lot of the methods that we’re building are very much focused on graph networks, for instance, because you can link different transcripts to different transcripts using different database connectivity, database pathways. We’re building foundation models as well on top of this, but also causal discovery. Using structural causal models to try and identify which particular feature in the network might be causing the underlying condition, as well as integrating a lot of genetic data into that.
Manuel: You need three things, right? You need data, you need compute, and you need people. So, tell me about these things.
Adam: Actually, the hardest challenge isn’t generating the data. It’s actually getting the samples in to generate the data. So, if you don’t have high quality metadata, you don’t have high quality patient material, then the data that you generate is often pretty poor. That’s the majority of public data, unfortunately. And we work in an area in musculoskeletal disease where there’s a critical data gap, we just do not have the data in the public sphere that we can collate and then really do important mechanistic machine learning or inferences on.
Manuel: Can you give us an example of what looks good? Because you’ve said it in a sort of abstract way. Give us an example of what should be good in terms of data and metadata.
Adam: I mean, if you just look at Geo, I mean, I’ve scraped so much public data over the years on many different projects. It used to be sequencing problems, because you had Solexa, you had Ion Torrent, you had Illumina. But actually that disappeared over the years because everyone’s standardised to Illumina, and the data quality just got so much better. Then you unfortunately have a lot of data relating to degradation of RNA, and that is still apparent within a lot of the QC, but most people don’t upload their patient metadata with their samples, so you have maybe some fantastic data, but you can’t link it clinically to anything that’s relevant. And so, because this problem is even more acute within the musculoskeletal field – I mean, if you look at the Human Cell Atlas project, it constitutes 0.4% of the overall Human Cell Atlas – you can’t train a foundation model. You can’t do really good mechanistic inference on that data. We need to generate that internally within the company, own that data, and then apply tooling on top from machine learning.
The reason why I think the next generation of biotechs need to be really, really crystallised in their biological anchoring is because algorithms are commoditized. They lose value because the good algorithms, everyone starts to use. So, basing a company on proprietary machine learning, for us isn’t where the greatest insights come from. It’s actually having high quality data that comes from high quality samples. We’ve even built our own internal biobank, HTA licensed biobank, inside the company to facilitate that. And then on the opposite side, because you can’t look at things in isolation, you have to work the whole workflow, identify the bottlenecks, fix those bottlenecks so they don’t become an issue, say, a year or two years down the line. So, the problem in our field, in musculoskeletal field, is that the animal models are just terrible. I mean, most drugs fail because the mouse biology is just so different. Mice don’t load weight like humans. I mean, they walk on four legs. They have paws instead of hands and feet, and their muscle has so many different cell types that don’t really phenocopy on top of the human muscle. So, we’re building internal organoid models, for instance, to solve that issue, so we get a mechanistic readout. And it’s that mechanistic readout that’s really important, because you can then build physical machine learning models on top of that, to then build a virtual image of a cell, to then give you some confidence that you know these targets are actually going to be useful when they get to the clinic.
Manuel: I think I’ve been asking most questions, so, one more question, and then I’m going to pass it on to Lyndsey. Is Oxford a good place to do what you are doing? And if so, why?
Adam: Oxford is fantastic because of the people. I love the fact that I can collaborate just across the road or down the hill. If I have a biological or a chemical or a physics problem, I can reach out to people, and I can find someone in the vicinity within 10 minutes. And there’s a really good collegiate atmosphere within Oxford. So that, for me, is why Oxford is a great place. And actually, the way Oxford’s set up is that it is very modular. I know different departments specialise in different things, so I often know exactly where to look if my problem is, say, a cell type specific problem. Or in blood, I would go to the Wynn, or RDM if I had issues with oncology problems. So, it’s almost like a directory system where you can find people, and people respond and are fantastic to collaborate with. I think that for me, is a defining feature that Oxford does really, really well.
Manuel: I was thinking about more entrepreneurial, capital risk. Because, you know, at the end of the day the UK could do a lot more if there was more risk capital and less aversion to risk. You’ve been mentioning risk. I think those sort of entrenched values, perhaps, how do you feel them in Oxford?
Adam: It’s a problem with the UK generally, access to capital. There’s a reason why most companies will leave the UK and go to the US, because there’s just not the capital to sustain a company after series A. Setting up an early startup is, again, really challenging. You have issues associated with that. Most universities take too much equity, and there’s plans to change that. Irene Tracy, Vice Chancellor of the University, wrote a report laying out some recommendations, so that’s actually worth reading. But again, it’s still challenging. There are some really, really good accelerator programmes, though, for PhDs and early scientists. Nucleate, for instance, is running a pretty good accelerator programme. But yeah, as I said, it’s still a challenging thing to do. I mean, it took me two, three years to actually push my way through the system, to get a spin out started.
I believe in founder led companies. I believe that if a founder has that ambition to drive their company, then they should be given the chance. And what the University in Oxford does very well is it has a more of a build model, where you have the likes of Oxford Science Enterprises who will inject capital to help form a company, but often they will bring in their own teams to then run that company, and it’s a different model. It’s a build model. It is what it is. And obviously, Oxford’s really benefited from OSE for a number of years, it’s why they have more spin outs than any other university in the UK at the moment.
Lyndsey: I could listen to you guys having this conversation for hours. You were talking about people before. And I think one of the big concerns that people have in this age of AI, is, if AI blows up and it’s able to do all these things, what happens to the people? And you seem to be very much in the camp of AI being more a tool to enhance what you’re doing. But how do you see the role of, say, the computational biologist, even the wet lab scientist, evolving over the next few years?
Adam: I’m not a pessimist. I’m very pragmatic in a lot of cases. And actually, I see a lot of opportunistic potential in the future. There’s a lot of doom mongering, and I get that, but I also know that AI has been here before. There’s a lot of hype and low cycles within the field of artificial intelligence, but whether we get Artificial General Intelligence, I don’t know yet. I suspect it’s not going to come from the current LLM focus, because we basically scraped all of the data out there from the public information from the internet, and these models have very limited ability. Reasoning was a fantastic achievement, but ultimately we need to move towards more reinforcement-type learning, it’s how children learn. You learn from your experiences, and you feed that back into the model. It gives you better ideas, it gives you better understanding. And we’re very far from that, I think, in the field as it is at the moment. But there are very clever people working on this. So, we might get there eventually.
But it’s going to take certain people’s jobs away, for definite, and we’re already seeing that, but it’s also going to lead to job creation. If you look at the history of mechanisation of farm machinery. Many, many 1000s of farmers were, unfortunately, back in the 17-1800s, were let go, but they all found jobs eventually, with the industrial revolution. So, it’s a realignment. I really believe it’s going to be a positive, have a positive impact on humanity.
Lyndsey: And what advice would you give to somebody who does find themselves in that position, where they have to maybe switch up what you’re doing?
Adam: I have retrained multiple times in my career. I went from working on rheumatology as a very basic scientist, working on one gene, one cell, CTLA4 and regulatory T cells. I knew that it was going to be problematic for my career if I stayed doing that type of research. I wanted to study more cells, in fact, all cells and all mechanisms, and I retrained, as I mentioned, in computational biology and machine learning. And I think that’s ultimately what scientists need to keep doing; keep iterating and making themselves better, learning new skills and not being too siloed. But having said that, there’s still a lot of science going on in assay development or skills that are very focused. But if you find yourself in a field where you’re struggling, I would recommend trying to look at retraining, or at least building skills. But I think everyone should be doing that anyway.
Manuel: I’m going to be a little bit of a devil’s advocate. It’s very easy to say, ‘retrain yourself’, and you’ve done it several times. But most people can’t or won’t do it. So, why can you do it and some other people cannot do it? What is it that you do that other people are not doing, in terms of retraining and always being current? What is it that you do, that you think is different?
Adam: The thing is, I don’t know whether that’s the case. I’ve got to put my hands up because Oxford can be a bit of a bubble, but I don’t know many scientists now who are in the wet lab and don’t know how to analyse data in R or Python. So, there’s already a drive for scientists to pick up these new skills, and I don’t see that changing. I think if you want to generate those really awesome scientific publications, then you do need a much more rounded background. And even research assistants in my group, even if they’re in the wet lab, will have to learn to code, and I teach them how to do that. So, finding a home is probably the most important thing that you can do as a scientist and finding someone to support you, because if you aren’t able to find that nurturing support, then no one’s going to push you to move out of your comfort zone. I think it’s a combination of internal drive, but also the environment that you find yourself in, and feeding off it. I don’t know what my life would be like if I hadn’t done my PhD at the Kennedy Institute, because I would never have been exposed to real translational science, and I wouldn’t have been able to really extrapolate and see how you could do that, and maybe my life would be completely different now, but I could say that about multiple steps in my life.
Manuel: Here is another challenge. I know you like challenges. You told me that. So, the other challenge is that most people are not in Oxford. Most people don’t have that amazing, nurturing environment. How can we become less dependent? I’m constantly thinking about this myself. I’m not in Silicon Valley. I’m in London. London is all right for AI, isn’t it?
Adam: I miss London. To be honest, I miss London. I love living in London.
Manuel: I love London too, but a lot of the time I have to carry my own environment. I have to surround myself with people who challenge me. And the way I do it is, you know, I have to become less dependent on the physical and become more dependent on the fact that we are all wired now, it is a network, right? A virtual network. What sort of advice would you give to people who are not at Oxford, who are not at Imperial, who are not at Stanford, so that they carry their own environment? How can we carry an environment that is nurturing? Because, as you said, we are willing, but if you don’t have the appropriate support, then you’re never going to be able to achieve as much. So, what would you tell us about how we can improve on that?
Adam: I mean, you slightly hinted at it in your question. It’s about developing a network and finding a home that is supportive. But scarcity drives innovation as well. I think it allows you to think in a completely different way, where you’re not just wasting resources on daft experiments. Actually, you think about and plan every single experiment to get the most information out of that experiment. How you do that is, I suppose, probably another two, three hour podcast. But the biggest thing is generating a network, and that doesn’t necessarily now need to be just a local network. How did I do that? I attend loads of different side events. The biggest problem is most of these side events are in London. So, you have to travel to London a lot of the time. But actually not necessarily now, you know, I’m actually originally from Newcastle. I’ve seen Newcastle change its ecosystem within the university substantially from when I used to live there, and there is a lot more emphasis on using omics approaches, and also spinning out companies as well within the north. There’s a new science park opening up all the time in the north. And actually it’s much cheaper science to do outside of Oxford and London, because the lab spaces are often a lot cheaper. The problem is the connectivity. The closer you are to individuals who are that like-minded, the easier it is, and that is why London is doing so well compared to the rest of the UK. It’s because it has that pull. And that isn’t necessarily driven by academic pull, per se. It’s the whole ecosystem that’s pulling people. You have DeepMind set up there, you have Recursion, you have Relation Therapeutics. There’s a wealth of different companies who’ve moved into that space, and they all have their own networking. They put on parties, and they put on side events and host the best speakers, and they can attract that talent, but you have to start somewhere. And I think I have seen, particularly in Manchester, for instance, a big appetite for building companies and bringing in more talent than I’ve ever seen before, over the past10 years, at least.
Manuel: How can we bring that crowd to the Festival of Genomics?
Adam: There’s a reason why the Festival of Genomics is held in London and not in Manchester. That’s because of the pull. You know, you will get people to London because a lot of people are already there, and people will travel. It’s really hard, and I don’t think there’s any one particular solution, and it has to also be political. And the biggest problem is, at the minute, the government has no money to really put into any of this. So, they want this, but there’s no mechanism at the moment for doing that. I think there’s a lot of talk about Mansion House tax reforms for pensions and that there’s going to be a flood of wealth coming in that’s yet to be seen. I don’t know, because I’ve never seen a plan on how that money is going to be allocated, spent, or actually invested. I know I’ve met a lot of pension fund managers over the past two years. None of them would ever make great biotech investors. They just don’t understand the market. So, we’ll see. Hopefully the government do have a plan. I just haven’t heard or seen any of it yet.
Lyndsey: You were talking earlier about retraining and doing a bunch of different things. And I had noticed when I was reading up on your research, you’ve done so many things. You’ve done epigenetics, single cell sequencing, computational biology, there’s so much stuff you’ve done. Do you have a favourite thing you’ve worked on?
Adam: Do I have a favourite? It’s probably these two, and actually they’re both not published. One of them was, in the early days of CRISPR, generating a guide RNA database for every single part of the human genome. And the reason why that sticks out in my mind is because it was a really complex computational task predicting which guides would be more effective, and then doing, at the time, very basic machine learning. But also, I managed to break pandas in Python, because there were more data points than could be physically contained numerically in pandas. And so that was like, ‘Oh, wow, I’ve actually come up with a problem that I need a completely different solution computationally to solve,’ and that was by sticking everything in a database, SQLite database. It also made me realise that actually a lot of old technology is still relevant today and still does a fantastic job. So, don’t just get hyped up on new technology. I’m a technologist at heart. I believe that technology is going to solve many of humanity’s future problems. But there’s also a load of old technology. Western blots are still important, qPCR. Do not neglect them.
The other one is actually preeclampsia. It’s essentially hypertension during pregnancy. I worked on this project where we were able to identify these tRNAs within preeclampsia. And the whole field is fixated on microRNAs, but we found some really interesting findings within tRNAs, and how these then block the ribosome and then basically kill endothelial cells, and that potentially is one of the reasons for stressing out the maternal system. So, two really random projects, but I have worked on quite a few things, butterflies, horses, humans, mice, rats, pigs, you name it.
Manuel: Another piercing question. Sorry, I need to keep my reputation up. I want to explore your love, hate relationship with the publishing process a little bit.
Adam: Now you’re going to get me started.
Manuel: Well, that’s what I wanted. It took just 50 minutes to get to that point. So, what do you hate about publishing, and what do you love about publishing?
Adam: We’ll start with the negativity, because I always like to end on a positive. So, the negativity. If you created a new field called ‘Science’ and you wanted to disseminate that information, how would you do that? The first thing to come to your mind wouldn’t be, hey, I know, I’ll stick everything in a PDF so people can read it. That just would not be a thing that you would do. You’d create a system where you could have interactive plots, where you could actually see the raw data. You could understand experiments a lot better by taking that information out and then starting to play around with it yourself. But at the moment, our primary source of information is by downloading a PDF and just reading it and then trying to interpret the figures yourself. The second thing is the fact that it’s a cash cow. It’s the only publication, media source that actually generates money, still, and it generates money from, unfortunately, from public money that should be spent on science and also charities. So, there’s this perverse nature of publishing where you know your career is basically dictated on where you publish, and not the quality of the science. That’s the negative feeling for me at the moment. The best thing to happen in probably the past 10 years is bioRxiv and the re-imagining of pre-prints and disseminating papers quicker. [Before], you would wait for two years before a manuscript gets reviewed. That science is then old, and so it just slows down the process of science. And they’re gatekeepers.
I said I would finish on a positive. Why do I like publishing? Well, it’s, ultimately, to disseminate knowledge. I always want to see impact in anything that I do. I think I’m driven by impact, and the transfer of knowledge is probably the most impactful thing that I think you will do in your career, whether that’s a teacher in a school or Professor at a university. I think knowledge dissemination is incredibly important, and that’s, for me, the reason why I still write papers and publish, even though I have sort of technically, personally given up once it’s on bioRxiv, but I know that my students and postdocs depend on me getting it into a high impact publication, whatever that means.
Manuel: We’ve gone on to the second hype word on my list today, which is impact. The first one was agentic AI. Now it’s impact. Everyone talks about impact, but everyone means different things. What do you mean by impact?
Adam: Impact, for me, is positively influencing someone’s life in whatever way. It’s very nebulous, and it’s very hard to quantify, and you can’t quantify it. Impact for me, just gives you a good feeling, and that’s how I internally quantify it. But I know there’s a lot of work within the research community of how you measure impact. It always revolves around numbers, unfortunately, and I think it’s a distraction. It’s not really quantifiable. So, I don’t think you should be quantifying it. And it has to be looked at on an individual basis. And I know, you know, the REF 2029, is going to come up, and it’s all going to be about impact, etc, but I think as a field, you’re going to really struggle to quantify that.
Manuel: Just for those who don’t know what REF is, it’s Research Excellence Framework, which is the system that, in the UK, your scientific output is measured by. And depending on how well you score on different things that are included in the final score, research, money, etc, then they give you more money.
Adam: To your department, not to your research group, though.
Manuel: Exactly, yeah. So, the whole academic environment is completely dependent on that. We’re already thinking about that at my University. It’s kind of important for those of us who live off science. Now, I want to go back into the AI thing, because you said biology first, AI everything. Well, that sounds a little like a contradiction to me. Can you explain?
Adam: It’s more about how you’re thinking. You’re thinking at every single stage, how you can automate a process, how you can make your life easier? You might not necessarily want to automate something or use machine learning at a different point, but it’s trying to think about where it’s actually going to improve your life. Because it goes back to what I said earlier. You have a finite number of hours in this world. You have to really utilise those effectively, and that’s often, for me, to make high level decisions, not low level interpretation. And if you can use machine learning to compress that down, then ultimately it’s going to benefit you as a person, because you can spend more time on things that really drive value, rather than things that are not that impactful. It’s always thinking about those particular principles.
Manuel: I want to think about the future. Do you think about the future much?
Adam: It depends. I used to think about the future a lot more when I was younger. I mean, now I think about it a little bit more, only because I now have a child, and I think that kick starts your brain into thinking. But actually, in terms of work, I have a vague idea of where I’m going. I don’t plan things out for years at a time, unless I’m writing a grant or something, and you’re forced to do it. I think in three week, four week schedules of what is actually the top four things that are really important for me to do today, because certain students might need your help. Certain problems need to be solved. You have to triage those and work out, ‘do they need to be solved right now?’ Or can I wait for an extra day or two for those to be solved? What’s the bottleneck?
Manuel: In terms of the technology, let’s sort of define that. What are the big hopes that you have in terms of technological achievements, and then what are the big anxieties that you have, let’s say, in the next five years? Because I don’t know about you, but the way I work today is very different even to how I worked six months ago. I’m not sure if that’s happening to you, but it’s certainly happening to me, the acceleration of the pace at which things are happening, and that gives me a lot of anxiety. I’m sort of answering the question by myself, but what gives you hope and what gives you anxiety, in terms of how technology is advancing, and this exponential growth on some aspects, at least?
Adam: For my field, it’s moving away from gene centric-based measurements, because the whole field is fixated on measuring genes. Ultimately, that’s a product of the fact that we have had technology that has only allowed us really to investigate gene expression at that level. But actually, when you think about the cell, it’s really complex. You know, your underlying transcript, your gene is compressed and comprised of your transcripts. And actually we often have not measured that, and that’s because we haven’t had the technology. But you know, long read sequencing that has led to us overcoming that inability to measure transcript expression effectively. But if you look at modifications on the DNA, for instance, we still have very crude ways of measuring methylation or measuring histone acetylation, deacetylation, and they’re all typically 1980s antibody-based approaches. And I really hope that in this area, there will be some improvements around those particular measurements.
Manuel: I mean, we have the central dogma, basically. So, what new dogma do we need to have? We need to have some kind of reference. Perhaps one could deduce from what you said that the central dogma is so outdated, even though we have these new loops coming up and down. But how do we need to think about them?
Adam: We need to start thinking of ways in which to measure that complexity, and think about how we can use machine learning in a much more informative way to not only help us predict the features that are important in a particular context, but actually how we can use those insights downstream to solve biological problems. We talked about publishing, but one of the other things that really annoys me is we often use tools, and I’m thinking of single cell sequencing, not as a tool to understand biology, but ultimately to make a pretty figure so you can publish high impact papers. And I think we really need to start thinking about what the biological question is that you’re answering, and can you use an already available tool, and if you can’t, push the technology forward. We always need better assays, and I think that’s not going to change in the future. So, if you can combine assay development with answering a biological problem, I think that’s where you really see the benefit for the future, or at least that’s how I see the future. Hope within science.
Manuel: Seems to me we might need to have a second interview, but I’m just going to finish with my now usual last question and I’ll hand it back to Lyndsey. Adam, I’m sorry, I’m attending your funeral and you have a congregation there. Somebody is going to read your obituary. What would you like to be remembered for? What would you like people to say about you when you are not here?
Adam: Positivity. I want to be remembered for being positive, for helping people achieve, sometimes the expense of yourself. And I’m not going to swear, but not being an idiot, because you can have local influence and you can have global influence. And actually, I as a person get as much benefit from having a local influence as I do on a global level. And ultimately, it’s a selfish reason. I like being the person I am, and I like being positive because I would hate to be negative and miserable, because I’ve worked and lived with enough of those people to realise that it puts you in an early grave. It’s very stressful.
Lyndsey: One final question for you, Adam, and hopefully it follows on from the positive side. It goes back to something you said earlier about it being so important to find your home and the community. What does it mean to you to be a role model, a mentor, to be part of that community, for someone else?
Adam: Everything, really. From a selfish reason, it gives me a lot of my own positivity, it comes from other people’s positivity. So, there’s a very selfish reason for it, but I think really, it’s probably just because I’m selfish! I like to be in an environment where people succeed and I want others to succeed. Now I’ve managed to develop a community around myself by being in a really great place, by interacting with some fantastic people that I’m always happy. Okay, not all the time, but most of the time.
Lyndsey: I think we can all be miserable sometimes, but sometimes it’s fine!
Manuel: So, if people want to follow your work and learn more about yourself, where would you direct to them to?
Adam: I need to update my website, which is my personal website, because everyone’s academic website is always out of date, and so is the university one. I think after this, I’ll probably update my website. You could find me very easily just by typing on LinkedIn. I’m on LinkedIn as well.
Lyndsey: Well, I think this has been such an amazing chat. I could have listened to you guys talk all day. Thank you so much, Adam.
