The Festival of Genomics and Biodata is just around the corner, and we’ve had the opportunity to sit down with some of our expert speakers to get a sneak peek into what they’ll be discussing at the event.
This week, we sat down with Omer Bayraktar (Group Leader, Wellcome Sanger Institute), to explore how his team is using cutting-edge single-cell and spatial genomics to decode the cellular complexity of the human brain. Omer shares how this technology is reshaping our understanding of glioblastoma, why tumour heterogeneity remains a central challenge, and how integrating spatial context can reveal entirely new biological insights.
Please note transcript has been edited for brevity and clarity.
FLG: Hi everybody. With the Festival of Genomics and Biodata just around the corner, we’re lucky to get the opportunity to chat to some of our expert speakers, to get a sneak peek at what they’ll be discussing. Today, I’m here with Omer Bayraktar, Group Leader at the Wellcome Sanger Institute. Omer, how are you?
Omer: Very good Lyndsey, thanks for the invitation. I’m excited to talk about our work.
FLG: Thank you for making the time to talk to me today. Could you start by telling us a bit about your career and what it is that you work on?
Omer: Sure. So, as you said, right now, I’m a Group Leader at the Sanger Institute in Hinxton near Cambridge, and I have a long-standing interest, over the past almost 20 years, in exploring cell type diversity in the nervous system. And really the gist of it is that there are lots and lots and lots of different types of cells in our brains that really lay the foundation for our complex abilities as humans – our cognition, our emotions, our behaviour. They all basically impinge on this rich diversity of different types of brain cells that we have in our heads. But that’s also a double-edged sword, in a way, in that diversity also exposes us to lots of different neurological conditions and diseases as well. So, our lab’s long term mission is to really understand the diversity of these cell types in the human brain and to explore how all those different cell types are vulnerable to different human brain disorders.
FLG: And what is it that first drew you to that area of science?
Omer: It took me a long time to get here in a way, to do the kinds of things that we do, where we use state of the art genomics technologies to study the diversity of cell types, and the molecular diversity of cell types directly in human brain tissue. My interest in this spans back to my PhD, many moons ago, in which I was trying to understand the basic principles of brain development using the fruit fly model, Drosophila melanogaster. And when I started out, my interest was really in what we describe now as fundamental biology, right? So, something that’s not directly connected to biology of disease, not studying human brain tissue. It did not involve any of the fancy genomic technologies that we use now and that we’ve used over the past decade. It was simply trying to understand the fundamental biology of how the brain develops, and how you end up with all these diverse cell types, from a small and seemingly homogenous pool of stem cells and neural progenitors early in development. And as I moved on to the further stages of my training, did my postdoc, and went on to start my lab, I think we have stuck to our fundamental enthusiasm about the diversity of cell types in the brain, and we’ve looked to basically apply that view and understanding of fundamental biology to human brain diversity and human brain diseases, and research areas that you might describe as disease focused or translational biology focused. So, I think that connecting the dots between the fundamental biology of how the brain develops and how genomics can help us identify cellular vulnerabilities and pathological processes in patient derived brain samples has been really the key in making this work.
FLG: Your talk at the Festival is about dissecting glioblastoma tissue with single cell and spatial genomics, and my, albeit probably limited, understanding, is that glioblastoma is also very heterogeneous. Now, first of all, why has that been your target? Why have you chosen to work on glioblastoma?
Omer: I’d say there’s two reasons that we focus on glioblastoma. The work we do in brain tumours is now about 50-60% of our focus, and a lot of that has been glioblastoma. The first reason is simply the unmet clinical need, right? Glioblastoma is one of the worst cancers out there. It’s an adult brain cancer, and unfortunately, it’s rapidly fatal and incurable. People tend to get it often in their 50s and 60s, and unfortunately, over 90% of the people who have glioblastoma will live only a little over a year after their first diagnosis. So, really, there is this unmet clinical need and patients are desperate for new treatments. And really, if you want to cure this disease, what we need to do is to understand its fundamental biology, in a way different to how we’ve been studying it over the past 20 years, and come up with basically new treatments ground in the biology of what’s going on.
The second reason is that we think it’s also a really fascinating use case to explore these high dimensional cell atlasing technologies, to explore how single cell and spatial omics can measure, either in single cells or in spatially resolved cells, the expression of hundreds of thousands of genes. We think it’s a really fascinating puzzle for those technologies to systematically identify the biology of disease. And glioblastoma, I’ll say, despite this intensive research that’s gone in to it over the past 20-25 years, with every single molecular omic technology that you could imagine, we are still blanking on the basic rules, the mechanisms that underlie this disease, and this is often what I describe as the ‘rules’ of this disease. The way we saw this was also a great challenge, and in a way, an opportunity for the single cell and spatial discovery methods to explore what’s going on in patient tissue samples in this cancer.
FLG: It sounds like it’s quite a complex cancer to study. What have been some of the biggest technical challenges you’ve had in this work?
Omer: In the field of glioblastoma research, perhaps the biggest challenge that many people could agree on is tumour heterogeneity. It’s this understanding that from one person to another, these tumours could be quite different. They can have different genetics, different mutations that underlie these tumours. They can have different cellular states, transcriptomic, metabolic or proteomic states that distinguish cancer cells from one tumour to another. But also they have extensive heterogeneity within each tumour. In a given patient’s tumour, the different cancer cells, the malignant cells, are not a monolith. They are not all the same. They actually exist across a spectrum of different molecular states, or cancer cell states. Actually, one of the things that got us interested in glioblastoma biology, that we perceived as a challenge at the beginning, was this understanding that in each patient’s brain, these cancer cells do indeed exist across the spectrum of different cell states. The challenge that we saw in the field was that before our work, the prevailing view of how this worked is that cancer cells have restrained plasticity to switch across these states, right? So, by using animal models and in vitro models, experimental work has suggested that a given cancer cell can go from state A to B to C to D. And really the idea was that that could happen in all kinds of unpredictable ways. Another cancer cell can then go from D to B to C to A, for example, right? We thought that this level of heterogeneity, this level of plasticity, while we acknowledged that it was a challenge, we thought that it was actually not quite the problem in itself. We thought that talking about the extensive heterogeneity of these tumours was, in a way, almost like an excuse. We thought, perhaps the real problem here is that we don’t understand how this works. It’s not that the tumours have a lot of different cancer cell states, but we don’t know enough about how cancer cells in a patient’s tumour behave, to be able to postulate whether they follow stereotyped cellular trajectories or just go all over the place.
FLG: That’s really interesting. And your talk is focusing on leveraging both single cell and spatial genomics. How do you balance using those two technologies in your workflow, and what are the benefits to using both of those approaches?
Omer: Our lab, I think we are known for primarily the work we’ve done in spatial omics. We’ve been in this field almost since its inception, and we’ve used all kinds of different spatial omic technologies, and we develop computational tools to make sense of spatial data. But I think our past and our future, which we share, I think, with many other researchers, is multi omic. We want to leverage all the different genomic technologies that explore these complementary sides of disease biology to maximise our discovery potential. In a lot of our projects, broadly speaking, we start with single cell omics. Whether that’s just measuring transcriptomes, or measuring both the epigenomes and the transcriptomes in individual cells. With single cell methods, we use that as a starting point to basically create a list of parts of that healthy tissue or that diseased tissue, like a brain tumour. And we start with that, because with single cell methods, where you dissociate a piece of tissue into cells or nuclei and then profile them, that’s still the highest throughput, most scalable method, where you can measure the highest number of things in each cell. So, we use that as our initial discovery engine, to be able to identify the disease-relevant cell types, cell states, or the gene expression programmes, or the epigenetic programmes that underlie disease cell states from patient tissue. And then with spatial what we tend to do is, again, use a lot of different technologies, but the way we see it is that with spatial we can then take that single cell data and, by integrating with spatial omics, put those cells back into the two and three dimensional structure of those tissues. That then really allows us to look at the biology of that disease in situ, so in the brain and all other solid organs. These disease-related cell types are not just randomly flying around or floating around. They live in very specific tissue microenvironments.
Back in the day, when we started out, we were doing these types of experiments, and we were kind of coarsely mapping cell types to still pretty high resolution, but only 50 to 60 micron little spots, such as those you observed using the Visium assay back in the day. But now, as you and many of the attendees of the meeting will know, you can do this at single cell resolution with imaging-based technologies. I’m happy to elaborate, but time and time again in this project and many others, we have seen the benefit of putting the two together, where often they’re very complementary to one another, and you really only see half of the picture, or just a small part of the picture, if you’re just working with one data modality.
FLG: And how do you make sure that the spatial context you capture is biologically meaningful?
Omer: Yeah, it’s a great question. First and foremost, it comes down to sampling. You have to look widely, if you can. Meaning, if you’re working with cancers and solid tumours, and you have these biopsies, regardless of the cancer you’re working with, you need to consider the regional heterogeneity of the tumours and the multi-scale spatial architecture. And what that means is really the understanding that within a given tumour, different parts of the tumour can harbour different biology. There could be different cell types. There could be different tissue microenvironments. So, that regional heterogeneity could really confound your observations if you’re only sampling it in one site. So, if you can, look widely within that sample, don’t just stick to a tiny little needle biopsy. If you can, work with larger centimetre scale samples, if your tumours are that big. And the second thing is to do that across different patients. I think one thing that needs to improve in the community of spatial omics is access to high quality patient derived tissue samples, and then the scalability of the assays would allow us to go from just a few tumours to dozens of types of tumours. But I think number one is good sampling, [ensure] you’re seeing enough of a given sample, and that you’ve got enough of them from lots of different patients.
I’d say, what that biologically relevant scale is, there is no real consensus. And really, maybe there should not be consensus, because I think the first question is, what is the biology that you’re interested in? So, if you’re interested in whether different molecularly distinct cancer cell states live in distinct tissue environments or in specialised tissue niches in a tumour, then of course, you have to look ideally at a large area so you see enough of a given tumour. And in our work in glioblastoma, this is precisely what we saw when we started looking at centimetre scale specimens, we were able to see that different parts of these tumour biopsies harboured those transcriptomically distinct cancer cell states. But then, your question could also be quite defined and specific. You could ask a question about how infiltrating immune cells come out of vasculature or leaky vasculature and stuff. And for that, it could be sufficient for you to just focus on a pathologically defined area of vasculature in a sample, so a small area where you do a refined type of analysis.
FLG: That makes a lot of sense. To get a little less technical, how do you envision your findings translating into clinical practice? And are there any particular barriers to that that you can think of?
Omer: Absolutely. First and foremost, in the type of work that we do, and many of our peers do, I think the most important thing that will come out of it with these technologies is the basic understanding of what the disease related cell types, the tissue niches, and the processes are. Because I think a really big problem in drug development and treatments of these complex diseases is that we don’t really understand what we’re dealing with. Of course, we want to have drugs that are working. And drug development takes a long time, so we need to get started. But target identification really requires us to understand the fundamental biology of these diseases. And ideally do that in patients, not using an animal model or in vitro cellular model as your primary source of evidence, but starting with an understanding of what those processes are in people, in their tissues. I think that’s really the critical starting point.
I think the second thing is, often with the kind of data that we generate in cell atlasing studies, you’re making this incredibly high number of measurements. To give an example, in our Glioblastoma Cell Atlas, we profiled over a million cells with single cell technologies. And in each cell, we measured thousands of genes on average, 4000 or 5000 genes in single nuclei. That’s like 4 or 5 billion measurements that we have made. A big challenge there is, how do you take that complexity and how do you whittle that down into a small number of actionable drug targets? A big problem in the field is that even though we have these new, cool, high throughput technologies, even though we have incredibly large data sets collected over the past decade, we are still not great at discovering the full extent of molecular targets, and a lot of the work still kind of just goes back into the usual suspects, such as previously studied pathways in those diseases. So, I think this challenge of taking these cell atlases and functionally screening targets to come up with some new targets for drug development is a big challenge.
FLG: Beyond glioblastoma, do you see your approach being applied to other cancers, either in the brain or elsewhere in the body? And if so, are there any particular cancers that would maybe be best suited for this approach?
Omer: Absolutely. I think the work that we do, the combination of single cell and spatial, and we’re talking about solid tumours here, I think this approach is relevant to every single solid tumour, cancers of adherent issues. We’ve sort of demonstrated this. In our lab, our work is, like I’ve told you so far, on glioblastoma, but we’ve already started to study other brain tumours, other types of glial derived brain tumours, such as astrocytomas, where there is also, again, an unmet clinical need. There are many other cancers in the rest of the body, from pancreatic cancer to colon cancer to lung cancer, where it’s known that these cancer cells are heterogeneous. They’re not a monolith of the same cell states, they actually exist across a spectrum of different cell states. They have that plasticity, and understanding that plasticity is important to see if cancer cells use that as a mechanism to evade treatment. But I think beyond that, this kind of roadmap of using single cell and spatial omics as a discovery engine of the basic rules of human diseases is broadly relevant to many different diseases, and many different organs.
FLG: What do you think is the next big thing that’s coming up, be that in the spatial omic space, or in neuroscience or cancer research. What do you think is coming next?
Omer: I think from a technical point of view, with spatial omics, there have been two important factors that have made spatial omics the field that it is now, and have taken it from a niche technology 10-15 years ago, and turned it into an omic technology that’s widely usable. The first is an increase in resolution. When we first started, the smallest piece of tissue that we could measure was actually rather large. It contained tens to hundreds of cells. So, that did not really give you cell type specific single cell biology. And now jumping forward, we are able to measure the expression of hundreds to several thousands of genes at single cell resolution with technologies like Xenium, MERSCOPE, CosMx, so on and so forth. I think that resolution has been great.
The second thing that puts spatial omics on the map, and the area where I’m expecting the biggest developments, is the multiplexing. The number of different measurements you can make on the same sample. So, how many genes can you measure? And I think there’s still work to do. At the moment, really, the state of the art – the number of genes that you can measure in large enough samples in a reasonable amount of time and a reasonable cost – for spatial transcriptomics is still a few thousand. This needs to really go transcriptome-wide, to at least 15-18,000 protein coding genes. And I think this then would get in the domain of profiling single cell RNA sequencing, but with spatial resolution. I think the type of information you’d get from that would be paramount.
Another thing is different modalities, of course. So, we study RNA because it is the modality that is the most convenient to measure at scale. You can measure, at single cell resolution, 5000 genes in a multiplexed way. You can’t do that for proteins. You can’t do that for lipids. You can’t do that for other omics. So, I think especially at the level of proteomics and at the level of lipidomics, pushing the multiplexing ability of those technologies, and thinking about how to put them together with these rich transcriptomic data sets, would be huge.
A sort of half technical, half biological thing is to think about how we analyse these data sets. If you think about spatial omics, for a while the attraction was that you’re looking at cells in situ, so you can see cells neighbouring one another, that are putatively interacting, that are forming these tissue microenvironments. But it’s really, really important that we don’t just base that on prior knowledge like ‘immune cell sits on a vasculature’, and so on and so forth, but that we think about unbiased discovery approaches that can be applied to this data to quantitatively define those relationships. I think there’s huge potential for various artificial intelligence/machine learning approaches that are being developed, to create what we can describe as foundational models of spatial omics. I think we are starting to approach the amount of data that’s needed, both in terms of all the different diseases or the tissues that are analysed, but also the number of cells or the amount of data sets that are generated from each one, to come up with foundation models that can quantitatively define what a tissue niche is and to describe at the level of genes – not just histological assessments or pathology – what these pathological tissue interactions and tissue niches are, and how they are conserved across different diseases, and how they’re different across different diseases.
FLG: If you had unlimited resources, let’s say money and funding is no problem, what experiment would you launch tomorrow?
Omer: You know, it’s a tough one to answer, because we want to do so much with this. But in my mind, the really pressing thing that we need to do is to take these cell atlases that we have generated, these multimodal cell atlases, and systematically translate those atlases into mechanistic, functional insight. To go towards drug development, to identify disease mechanisms. It’s not enough to say that an immune cell expressing gene A is near this cancer cell. We need to be able to actually understand what gene A does in those cell types, and we need to do it for hundreds and thousands of genes in different disease relevant cell types. So, I think the types of experiments I’d like to do, we’d want to take our glioblastoma cell atlas and use the most complex in vitro models for those tumours, such as patient derived brain tumour organoids, which I’ll talk about at my presentation, and do large-scale cell type specific perturbations in them. And connect the dots, so to say, ie. we did the cell atlas in the patient derived tissue, in the real thing, and we found this gene expresses in this location, then in our functional experiments, in vitro, we target that gene and we actually attach a functional consequence and mechanistic consequence to that.
FLG: That does sound really interesting. Now I’ve got a couple of final questions for you. As you said, you’re speaking at the Festival of Genomics and Biodata in January. What are you looking forward to about the event, and why would you encourage people to come along?
Omer: The Festival has been going for a little while, I think this might be the third or the fourth time I’m attending. I’ve kind of like lost count of it. I think the Festival has done a good job of nurturing a community and becoming a regular, recurring event for genomics people. It’s good to see, it’s good to have this event and to really catch up on all things genomics that are going on in the UK, and to interact with people from diverse areas of academia but also diverse areas of industry, you know. There’s a strong pharma R&D presence there. And it goes without saying, there’s lots of international speakers who show up there as well. I’m looking forward to catching up with some old friends and making some new ones, and hearing what’s going on. I’m biassed, in that I like spatial omics a lot, and this year is the first one where there’s a spatial stage, so I’m quite excited to participate in that. And I think there’s some really cool speakers, including various UK speakers, but also European and US based speakers that are coming in. I’m definitely looking forward to that as well.
FLG: Well, thank you so much for your time today. It’s been a really interesting conversation, and to everybody watching, I would take Omer’s advice and come along to check out our new Spatial Stage at the Festival in January, where you can hear much more about what we’ve been discussing today. Omer, thank you again.
Omer: Thanks so much, Lyndsey, it was a pleasure.
WANT TO HEAR MORE FROM OMER BAYRAKTAR AND 360+ EXPERT SPEAKERS? REGISTER FOR THE FESTIVAL OF GENOMICS AND BIODATA NOW
