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.
In today’s interview, we speak to Matthew Loose (Academic Lead of DeepSeq, University of Nottingham) about how real-time sequencing could transform cancer diagnostics and care. He explores the evolution of long-read sequencing, from early innovation to its emerging role in cancer diagnosis, and reflects on the challenges, breakthroughs and collaborative efforts shaping a future where patients may no longer wait weeks for critical answers.
Please note transcript has been edited for brevity and clarity.
FLG: Hi everybody. The Festival of Genomics and Biodata is just around the corner, and we’re really lucky to get to spend some time with some of our expert speakers in the run up to the event. Today, we’re sitting down with Matt Loose from the University of Nottingham. Matt, thank you so much for taking the time to speak to me today. Could you just start by telling our audience a little bit about you -your background, your career, what it is that you work on?
Matt: I am a Professor of computational biology and developmental genetics at the University of Nottingham, and I have been here for, oh, more than 20 years now, leading a team to address different questions over the course of that time. But where I’m at now, we are using cutting edge genomic tools and techniques to address questions very, very quickly. We’re really interested in how we can help patients answer questions that are important to them within the shortest possible timeframe, and then how that can also help clinicians in managing the treatment of patients. So, we’re working in lots of different areas. The work I’ll be talking about at the Festival of Genomics is in the cancer biology space, and that’s where we’re seeing the most impact right now,
FLG: And you’ve stated that you have a ‘long-standing interest in computational biology and its intersection with developmental biology and genome structure.’ So, how did your research path then evolve towards cancer diagnostics and rapid sequencing workflows?
Matt: Yeah, that’s a great question. So, back in 2010-2012, around then, we were very, very interested in looking at the genomes of non-model organisms. An awful lot of the work that we do day to day in science is looking at organisms that are not well studied, and if you don’t have a genome for those, then it makes looking at the organism more complex. And so back then, I was looking for new methods that would allow us to more easily sequence genomes of organisms that people just hadn’t looked at. And the method that we picked up on in early 2014 was Nanopore sequencing, this ability to sequence very, very long molecules. And getting access to that technology was transformative to the sorts of things that we could do. Very quickly, we went from looking at non-model organisms to actually exploring the technology itself and saying, Well, what’s possible if we manipulate the sequencer to do different things? What can we do with data that can be analysed as soon as they’re generated, or even whilst they’re being generated? And we probably spent five or six years developing those methods to look at data in real time, to interpret results as they appeared. At that point, that was all largely theoretical. Then in early 2020, of course, we all went through that period of lockdown and being exposed to all that was happening, and actually, we were using the technologies that we’ve been working on to study the COVID virus. So, we were part of the COG-UK Consortium, and we were contributing sequences to that. That gave us much, much better access to clinicians and the ability to work with clinicians and build relationships with them. It was 2022 when we first started working in earnest with teams who were looking at cancer to try and give rapid diagnoses, and to try and tell patients within 24 hours, or actually during the course of an operation to tell clinicians what they may be dealing with. It’s been quite a long journey, where a biological question drove us to develop new methods, and then we realised, well, hold on, these new methods can be deployed in all sorts of new ways. It’s been absolutely fabulous to have the opportunity to do it actually
FLG: What are the benefits of using long reads specifically for applications like this? And did you face any challenges when adopting those solutions in your research?
Matt: The interesting thing about long reads in the particular application that we are using is, you don’t really need long reads. It’s a bit of a contradictory answer, but what you really need is the ability to look at the data as soon as they’re generated, and we don’t really care about the length of the molecule. Now, there is something that we can do as a consequence of using long reads, and Nanopore sequencing in particular, which means that we can select the molecules that we want to look at whilst we’re sequencing. We can choose to look at some regions with long reads in a lot of detail, and other regions, we can just measure tiny fragments of the reads and then choose others. So, the key benefit for this application is that Nanopore long read sequences are genuinely real-time. You can analyse the data as the molecules pass through the Nanopore, you don’t have to wait until the end of sequencing to run your analysis.
Did we face any challenges? Gosh, well, we’ve been doing this for a long time, well over 10 years, and there were all sorts of challenges when the technology first launched. Yes, it was incredible, absolutely mind blowing, but the accuracy of the reads was perhaps not sufficient for us to be able to do the sorts of analysis we would want to do, and so we’ve had to wait for the accuracy to catch up. And that has happened. It happened years ago now, that it’s good enough for us to use. We also had developed ways of looking at molecules in real time, because nobody had done that before. So actually, every piece of the puzzle was a first step. It was the first time that somebody had tried doing that. So, yeah, the benefits are significant.
The other thing is the sequencers are incredibly portable and small, so we don’t have to centralise them to one location. We can put sequencers close to patients. There’s all sorts of possibilities and opportunities. And it’s almost like one after another, you’ll think, right, we can do this now, how are we going to solve the problems associated with this and keep stepping through?
FLG: It sounds like there’s a lot of innovation that comes into making those decisions. And how do you balance the innovation between the wet lab side of sequencing and the, really important, computational side of it as well?
Matt: That’s a great question, and both of them are vital. The wet lab side is key. If you don’t put good material into your sequencer, no matter what computational tricks you have at the other end, you’re not going to be able to do what you want to do. Actually understanding how wet lab processes work, particularly as you get closer to the clinical side – the precise pathways that samples go on, who’s involved, how they’re going to interact with a sample at a given moment in time – is probably as important as the computational stuff. Getting both of those right is absolutely crucial. How do we balance those? It tends to be that there’ll be moments when one is more important than the other. I would say actually, now, the bottleneck for what we want to do is probably the wet lab side. And by that, I mean it’s training people to run the devices and the protocols in an efficient way, and from our point of view, understanding how people need to use it, whereas the computational side currently is solved for what we need to do. I would imagine that that will keep on tick-tocking from one to the other as we continue to develop.
FLG: Nanopore technology has changed the speed and the accessibility of genomic data, but what do you see as the next major breakthrough in that space, and what limitations might need to be addressed to make that happen?
Matt: That’s a very good question. Right now in the UK, we have a system of distributed centres for running sequencing, which is fabulous. Nanopore sequencing means we can make that even more distributed. But to do that, we need to address those training issues. We need to simplify the wet lab side. We also need to think very carefully about the data analysis side. And the interesting thing there is you can easily centralise data once it’s been generated, and that means that you can have expert data and analysis, and analysts looking at samples, wherever they may be, simply because they can be shared in the cloud. But we need to build those sharing processes. I won’t be doing this, this is something that has to be done within the NHS, I think, in terms of what’s next. I mean, the key challenge, I think, is rolling this technology out to people to be used.
I don’t actually want it to change for a bit. Sounds very boring, but I would like it to remain stable for a while so we can train people up and get them using it. There aren’t any changes needed in DNA sequencing on Nanopore at the moment. There are lots of exciting things coming down the line, Oxford Nanopore can sequence RNA, so we can look at RNA without copying it. From a research perspective, there are lots of things that we can find there. And there’s talk about people being able to sequence protein in the future as well. So, I think there are huge opportunities to come. But in the DNA space, we actually need a period of stability just to roll it out.
What limitations might there be to rolling it out? I think partly, whenever you change the way people do things conventionally, there is just a block to switching to a new method. ‘I’ve been able to do it this way for a long time, why should I change?’ So, as I think we are doing, demonstrating the patient impact that this can have, and motivating people to make that little bit of extra effort that’s required to switch to a new pathway, it’s those sort of practical, logistical things, like collecting the sample in the right way, and all those small pieces that need to be done, that are absolutely vital to make the larger piece of work succeed.
FLG: I think that’s a really great way to look at it. It’s all well and good to think ‘we could do all these amazing things’, but training people up in what we’ve currently got is really important. I think often when I ask questions like that, I’m expecting the really exotic answer, but I think that is so important to consider as well.
Matt: The funny thing is, we’ve had the really ‘exotic’ for 10 years, and actually we need to make it stop being exotic and start being normal. Then we can go on and try other things. But it’s almost like, there’s many people all around the place doing incredibly cutting-edge things with this, but they end up being so far from where the actual clinical need is that the gap to catch up becomes too much. We have to bring everyone with us, and hopefully that’s what we’ll have an opportunity to do.
FLG: And on the topic of clinical need, how does having sequencing data during surgery change decision making from a clinician’s perspective?
Matt: I should be clear, I’m not a clinician. I’ve just had the pleasure of working with clinicians and talking with them about what they can do. And I think not every case will be impacted by having the sort of data that we’re generated during a surgical time frame, but some will. The reality is you will have tumour types where the surgeon may consider a more aggressive surgery or a less aggressive surgery simply on the basis of what specific tumour type it is, and we’ve seen examples of this in our research-only use of the technique so far. For example, a patient who comes in, who the clinicians are unsure as to the exact brain tumour type they have, and we identify something called an ependymoma. Now, with an ependymoma, the treatment would be a significant resection of all of the tumour, and the surgeon will be quite aggressive. But where they’re not sure about that, they will halt the surgery, wait for the full molecular information, and then perhaps have a second surgery, with all the extra risks associated with a second surgery. And already we have seen cases where we knew in the first surgery, and we could have informed the surgeon during the first surgery. What we find is that there are cases where the surgeon may change the exact nuance of what they’re doing in response to the information we get.
Crucially, then, you have the information you can give back to the patient straight after the surgery or within a reasonable timeframe. Currently, patients may wait 4, 6, 8, 12 weeks to get the precise information they need. And in some cases, that will be important for selecting which type of treatment pathway they go on. It may be important for access to clinical trials and new drugs, and they’re waiting a long time for this. Our goal is to provide the data to the clinical team by the next multidisciplinary team meeting happens to discuss that patient. So, instead of that team having to wait weeks, they have everything they can get at the first meeting, which allows them to go back and talk to the patient, hopefully before they’ve been discharged from that round of surgery, to give them the information they need. So, actually, it has the potential to change things at the surgery step, but it definitely will impact treatment pathways and all of those extra steps that happen later on.
There was a paper published recently from a group in in Utrecht, and off the top of my head I can’t remember the exact figures of the number of surgeries that they would have changed, but it was a significant number where the surgeon said yes, this would have influenced our decision. Now, to what extent would you have to talk and discuss this with a surgeon? Ultimately, having more information is better than having none. We often have surgeons now ask for confirmation. You know, is the result back? Does it support what we think, yes or no? And working on how we can use that and deliver it effectively is going to be the next step that we need to do.
FLG: And your talk at The Festival of Genomics and Biodata addresses this in brain tumours. But do you envision this approach being adapted for other cancer types or even other disease applications?
Matt: Absolutely, I think so. We’re already starting to work on some other tumour types which may have similar needs. I think CNS tumours are a unique and very, very specific example of where everything comes together perfectly into this one assay. But the moment you see that you can deliver results this quickly, we’ve had other clinicians come knocking on the door saying, well, we need this. Things like soft muscle tumours and sarcomas are something that we’re actively looking at. We can pretty much use everything that we’ve been using with CNS tumours in those samples as well. We also have looked at leukemias, where we can provide a very quick and rapid diagnosis. We’re starting to look at things like ovarian cancers, where patients may wait eight to 12 weeks for results to come back, and that will be crucial as to which type of treatment pathway they go down. So, if we can get that to within the time they are in theatre or within the time they’re in hospital, that has far more significant benefits for them.
When we move outside of the cancer field, there are other examples where this technology is being used. Things like rare disease in newborns, where they may be very, very unwell and you want a quick result. Now, whether you would do whole genome sequencing and give a rapid turnaround, or whether you would do something targeted, as we’ve been doing, probably that would be up to clinicians to investigate. But yes, there are lots of different potential candidates for this as we get better and better at using it.
FLG: How do you see the role of machine learning or AI in interpreting real-time sequencing data and supporting surgical decisions? I suppose that’s the question everybody’s asking right now!
Matt: Yes, absolutely, a significant question. Ultimately, all interpretation of results has to be done by a clinician who is aware of the full range of information about the patient that’s in front of them. In our case, we’re looking at molecular information, but there will be many other things to be taken into account around the treatment of a patient. What I think about machine learning is it obviously makes it possible to do some of these things a lot faster than we have been doing in the past, to synthesise results and to present them to clinicians in a usable form. I don’t think we are ever… well, I don’t think at present we’re going to see an AI diagnosis. You’re not going to put these data into ChatGPT and it will give you an answer. We’ve all used platforms like that, and we know how they can go wrong. But working out how we can use these methods in a controlled and rigorous way is absolutely crucial, and some of the methods that we are using use machine learning approaches to classify tumour types into different categories, say, and they can do it incredibly quickly. So, I imagine that it will grow, but it will grow quite carefully and it’s always the clinicians expertise that that has to be the final endpoint.
FLG: If resources and technology were no constraint, what’s the one question about genome structure or cancer biology that you would like to answer next?
Matt: If resource and technology were no constraint, I would like to carry out methylation based profiling of every tumour type that comes through our hospitals, I’d want to know what the signatures were, because I think this is an under-investigated area in many tumours, other than the ones that people are looking at in depth already. I think [it would be useful] to have a large database of those methylation types, and to have that genome wide. Currently, when we look at methylation, we’re usually looking at microarray profiles, we might be looking at 450,000 sites or 800,000 sites in the genome. There’s millions, 20 million plus sites, that may be clinically important, and we haven’t systematically surveyed them. I think for diagnostics, for classification, this could be really, really significant, and it may allow us to see signatures of changes that we haven’t seen before. I couple that then with structural variant information as well. Knowing about the gene fusion events, chromosome rearrangement events, again, it’s very complicated to look at those with short read sequencing, but relatively straightforward to look at with long read sequencing, and I think that we have most likely missed many subtleties in that area, simply because we haven’t had the tools to look. So, yeah, to pick up methylation profiles and structural variants across all of these different disease types would be absolutely incredible.
FLG: It sounds incredible! Now, a final question for you, what are you looking forward to about The Festival of Genomics and Biodata, and why would you encourage people to come along to the event?
Matt: I was at the Festival last year in person. It’s incredible to see so many different groups all gathering together to talk about questions around genomics and biodata, a diverse array of topics. It’s an absolutely fabulous place to get everything that’s happening right now and to get an update on it, but also to then network with other clinicians, other academics, members of the public, the companies that are represented there, from whom it’s always useful to learn what they’re doing next. It’s a brilliant place, to have all of these different things in one space, and to be able to talk to everybody. So, I am just looking forward to meeting so many people again who are interested in these areas.
Why would I encourage other people to go? Well, I think it’s a great place to learn so much. It’s a busy day. If you stay for the whole time, and you interact with everybody, you’re certainly going to be worn out at the end of the day, but you will have learned a huge amount!
FLG: Yeah, that’s definitely true. There really is something for everybody on site!
Thank you so much for your time today, and we’re so excited to see you there on site in January. You’ve mentioned that it’s great opportunity to meet people, so I’d encourage anybody who’s listening to this interview to come along and meet Matt and hear more from him. Thank you again, Matt.
Matt: Thank you very much.
WANT TO HEAR MORE FROM MATT AND 360+ EXPERT SPEAKERS? REGISTER FOR THE FESTIVAL OF GENOMICS AND BIODATA NOW
