PROFESSOR CHRISTIAN SPECK combines a biologist’s approach with an engineer’s mindset, he’s worked his way to the frontiers of cancer research and cancer therapeutics by always looking out for new tools to accelerate his discoveries.
Speck grew up in Berlin amidst a family of scientists; his mother is a biologist, his father a zoologist and his middle brother is a chemist. His youngest brother is a police officer and Speck says he too sometimes plays detective in the lab as he investigates what happened in experiments.
Now a professor of genome biochemistry and molecular biology at Imperial College London, his lab was recently awarded Cancer Research UK’s prestigious Discovery Programme Award to explore how his breakthrough research on DNA replication can help fight cancer.
He visited InstaDeep’s headquarters to talk about his lab’s research and how artificial intelligence is helping push the boundaries of biology.
This conversation has been edited for length and clarity.
Karim Beguir: What inspired your focus on DNA replication?
Christian Speck: Mistakes in DNA replication probably cause many cancers. I am someone who’s always asking how things work and DNA replication is a fundamental process that goes wrong in cancer. It’s also one of the fundamental areas that you try to attack with cancer therapy. So it’s actually really close to treating patients.
It’s very timely in terms of impact. What do you see as low-hanging fruit for your research?
We discovered over the last 10 years how the replication machinery assembled. That was hugely exciting for us to understand how this process works. Now the low-hanging fruit is to use this knowledge to develop inhibitors. We now have tools in our hands to use structured information to develop structure-guided small molecules, plus we also have very powerful biochemical assays to screen for specific inhibitors. We have all the tools available to make the next step.
In the last 10 years, we were mostly focused on budding yeast. It was hugely exciting for discovery science. But if you want to translate that, then you need human tissue. We now have two PhD students working in this direction. This Cancer Research UK Discovery Award will allow us now to really progress full steam to work on human DNA replication.
Christian Speck
“They hold a lot of promise to influence the ageing
process in a positive way. I could see a revolution
happening in that area.”
So the process is, first, to understand the mechanism through which cancer can arise and through a granular understanding of this mechanism, potentially design cures or cancer therapeutics further down the road?
Yes and you also have to consider that many cancer therapeutics currently on the market are chemotherapeutics. Most of them target the DNA replication process in some way or the other. But they’re very unspecific. They damage the DNA. And there are hardly any drugs that target the replication machinery specifically. Maybe there’s one exception: the PARP inhibitors. They’re a famous class of molecules that have been a big success. Our aim is to add to those classes.
It’s like moving from carpet bombing – chemotherapy is very hard on the body – towards more targeted precision-style action against the cancer cells.
Yeah, this is exactly what we’re moving towards. Chemotherapeutics attack DNA directly. In this way, they frequently cause a huge number of mutations. If you sequence the genome of patients who have been treated, what you can see looks like you put DNA in a blender and put it back together in a random fashion. There is a huge amount of damage that gets done and multiple mutations. This is probably why many patients relapse, because so many mutations are generated that could impact other cells in the body.
When you say relapse and new mutations, this would be another cancer, but not necessarily the same as the initial one?
Yes, it could be, or it could be that some of the cells survived and then further mutated due to this event. With cancer, it’s frequently that they’re hiding in a state where they don’t grow. But in this state, they can still be damaged. The DNA can still be damaged and can still have additional mutations.
When I hear DNA damage, I also think about the impact in terms of finding solutions to DNA damage that could help longevity as well.
There are a few areas that are really getting targeted right now where you try to clear cells that got damaged. Those are usually the senescent cells. These are cells that something has happened to – damage or some other kind of problematic event – and they decide to shut down on themselves. And that’s senescence. The trouble with those senescent cells is they cause a lot of signalling events in the body; a lot of immune reactions. And so, this is really problematic during ageing. There are a lot of groups that now try to clear away those senescent cells. We’re still only learning how this process works. But, it could be that some of the molecules that might inhibit DNA replication could also drive cells into senescence, or kill particular senescent cells. We’re collaborating with a group that studies this process. They’re very interested in testing our future inhibitors.
I’m curious about the tools that you use and how they’ve evolved through your career.
I started off as a biochemist. But if you can see from only one angle, it’s not very satisfying. I expanded in the direction of more genetics and also chemical biology approaches. For a long time, I also collaborated in the area of structural biology and electron microscopy. We published some of the earliest electron microscopy structural work in the replication field. And then this became very dominant, because there was a structure revolution and electron microscopy became the most dominant technique to look at large protein complexes. This is when we then started to develop in-house capability for that.
When was that?
Around 2015-2016, I was in contact with the now-Nobel Prize winner from the LMB (Laboratory of Molecular Biology in Cambridge) and he advised me, so he was listed as a collaborator for our Wellcome Trust grant. More recently, we moved also into genomics. Why? Because now genomics is such a high resolution. So high resolution that you can combine structure information with genomic information and it gets a similar read out.
Now you can transfer some of your knowledgein vivo. And we’ve been doing this to look at the same process with high resolution cryo-EM or genomics, and combine this with proteomics to understand how the process works. With proteomics, you can ask who is involved. With cryo-EM, you can look at the shape and amino acids, and maybe you can get a mechanism and with genomics, an in vivo view of the situation, because you want to verify things at least in vivo.
This is how we play with our systems. The latest addition is now the move into molecular dynamics and AI. This is something I want to carry forward in the future. I think this will become more commonplace.
Karim Beguir
“AI-based methods hold promise because
they can work on very large data sets.”
Do you see research activity becoming more and more in silico – compared to classical lab technologies?
Definitely. Already on the cryo-EM and the genomics workflow, the lab-based work is maybe one third, and then there’s probably two-thirds of analysis that happens in respect of effort, or maybe 50/50. This will only expand because now if you’ve worked particularly in a structure area, you have parts of protein machinery that are well folded, and they can be handled by cryo-EM and so on. But then they move, and they do things, and they assume different structures; some of it can be handled by cryo-EM, but only some of it. Molecular dynamics simulation can help a lot with that.
When we look at, for example, molecular dynamics, which is basically, conducting full simulations like full physics-based simulations of large molecules and looking at their evolution, what would you say in terms of the quality of the results with the equipment we have today?
So far, from what we’ve seen, the data makes sense, and they’re good. They are heavily dependent on data processing and GPUs. And the good news is that this area is a technology that is developing incredibly rapidly. Every year, the processing power goes up. And so the quality of the data will only improve. They’re already good to very good now. But the potential is incredible. This is why I also feel, as a biologist, I need to invest in this area. It is the future.
Thinking about the future, for your work in cancer research, what do you see as ambitious but realistic breakthroughs in the next five to ten years?
It’s two-fold: you want to discover some new mechanism, which will probably produce some interesting papers, but we also want to test key reactions to see if we can develop good inhibitors, and then see how they affect normal cells and cancer cells. The idea is cancer cells should be hypersensitive to the inhibition of DNA replication. This has been tested in the context of molecular biology settings or cellular settings, but not with drugs. This is where we want to go and test specific aspects of the process. Maybe during this time, we’ll want to start a company in this direction to raise more funds to further extensive studies so that we can approach preclinical and maybe clinical work.
hat will be exciting. Do you see a path where we understand DNA replication so well that we could build cancer therapeutics, or would this be based on something more proactive before a disease develops?
First, the thing is probably therapeutics. If this can develop into something proactive – and there’s a chance – we’ll test it. But it’s less established by now. Scientists are always careful with their words.
“Whether it’s understanding language, or understanding the
Karim Beguir
language of biology, we definitely see an acceleration. We see
things that few people believed were possible
becoming possible.”
I work in AI. And what’s very clear from our point of view is AI is definitely developing extremely quickly. If you look at recent developments, they are nothing short of spectacular. Whether it’s understanding language, or understanding the language of biology, we definitely see an acceleration. We see things that few people believed were possible becoming possible. Do you feel the same about your field – and in particular, the development of science against cancer? Do you see that kind of acceleration where things that were considered completely remote are now starting to open up?
I work in AI. And what’s very clear from our point of view is AI is definitely developing extremely quickly. If you look at recent developments, they are nothing short of spectacular. Whether it’s understanding language, or understanding the language of biology, we definitely see an acceleration. We see things that few people believed were possible becoming possible.
Do you feel the same about your field – and in particular, the development of science against cancer? Do you see that kind of acceleration where things that were considered completely remote are now starting to open up?
Yeah, those developments in AI definitely helped to speed up some of the bottlenecks. In biology, there are still bottlenecks.
What are those bottlenecks?
One of the bottlenecks was the reconstitution of those processes. Now we achieved this with one part of the reaction process in DNA replication. Now this opens up a million experiments and a million more questions we can ask.
This is why this Cancer Research UK grant was given to us, because now it poses exciting opportunities. We do want to take full advantage of that and we’ll use all tools that are at our disposal. With all the structural work that we’ve been doing, but also the genomic work, there will be possibilities that we can take advantage of AI and machine learning and molecular dynamic simulations.
Karim Beguir
“Whether it’s understanding language, or understanding the
language of biology, we definitely see an acceleration. We see
things that few people believed were possible
becoming possible.”
Beyond cancer therapeutics, what would you say over the next five to 10 years could be a big break- through in biotechnology?
Ageing. That is the other area for us that is of interest. And also going further into the direction of epigenetic processes. Ageing gets misregulated. Information that is not genetic gets put on top of the genetic code. Why do protein systems that are associated with DNA, then become modified to those epigenetic modifications? They hold a lot of promise to influence the ageing process in a positive way. I could see a revolution happening in that area.
At the beginning of the 20th century, average life expectancy was about 50 years old. Now it’s around 80. Could you imagine seeing another jump like this?
A big jump was thanks to antibiotics. Those dealt with a whole class of common problems, like bacteria. Now, what we are fighting against is not one single thing, and if you can attack it, then you’re done with it. We are dealing with a series of person-specific problems. You need to develop something that deals with many different issues. This is why life expectancy is increasing in general – but not in the last three years. To help that kind of jump again, like with antibiotics, that seems very, very tricky. Because everybody has a different kind of issue, problem, or genetic makeup, and needs a different treatment.
That’s probably why computational or AI-based methods hold promise because they can work on very large data sets, personalised data for each person’s genome and the like. That could be helpful for the future.
Yeah, it’s too many variables, right? The human mind can hold seven variables. But there are more than seven variables that influence ageing or cancer. There’s a very clear reason to do a shift in our thinking and use AI technology to help us make sense of the huge amount of information that we have.
I would like to ask you a question, Karim. You’ve mentioned how AI might move things forward in the future. I’m always looking for the next tool that will allow us to make a big discovery. What is the next thing coming on the market now, or in five years, or 10 years’ time?
Christian Speck
“The human mind can hold seven variables.
But there are more than seven variables that
influence ageing or cancer. There’s a very clear
reason to do a shift in our thinking and use AI
technology to help us make sense of the huge
amount of information that we have.”
It’s a very exciting time in AI. First we’re seeing an explosion of data. And this is particu- larly true with biotechnology with the cost of sequencing a genome now being below $1,000. And at some point, hopefully getting to $100, which would unlock tremendous capabilities with AI systems. Perhaps something which is less mentioned is this is not just about more data and more compute, AI is progressing at the level of the science itself. It’s the algorithms themselves.
There have been recent studies which show that the efficiency of AI algorithms is following a sort of Moore’s Law. So every 16 months, the efficiency of algorithms – how much compute I need to get to a certain level of accuracy – has been divided by two. For me, that is exciting. It means that we are constantly inventing better, cleverer ways to use the compute and the data we have. There’s never been more smart researchers looking at AI in the world than today. I’m quite optimistic that we’re going to continue to see amazing, technological discoveries.
For me, the biggest breakthrough that AI could make in the biotech area is if you could simulate a protein function – its complex machinery – where you can simulate all steps of what it does. Say you have a cell in a computer, and you can start to simulate all the processes that can happen. This is the journey that you will take: it’s a cell, and then the next thing is an organ. And then the next thing is a human body, and actually understanding how the human body works. And then you can maybe also look into, or try to understand, what ageing really does because you need to consider the whole human body in this context.
Absolutely. And we’re starting to see this. At InstaDeep, we’re working on DeepChain, which is all about designing proteins for specific tasks. We’ve seen amazing breakthroughs by the research community around protein-to-protein interaction, 3D structure proteins. I totally agree with you that the next step will probably be moving from a single interaction to multiple complex systems. And hopefully, from there, we can scale to understand the multiple pathways and ultimately the cell. It will take time, but I believe that it is possible and exciting. I could imagine many breakthroughs along the way.
This interview was recorded in 2022.
