DNA graffiti: mutation patterns in cancer | 91TV
Transcript
- Thank you to the Royal Society, of course, for this immense privilege and honour to be standing
- here today and sharing with you the work and the journey that we've taken in exploring mutations
- in cancer. So, many in my team are here today, and many of you have been alongside us in our
- research endeavours. Thank you. I speak today on behalf of us all. So I'd like to start by
- first acknowledging that we are standing on the shoulders of giants, the giants of DNA,
- biology and chemistry, the visionary leaders that thought about the Human Genome Project,
- completed it and shared data. All that enabled the development of modern chemistry techniques
- that accelerated our ability to read the human genome, and that's DNA sequencing. Without
- that prior context, we cannot have derived the insights that you will hear about shortly, for in
- the last few decades, the speed and scale of DNA sequencing has increased by orders of magnitude.
- Thanks to the advent of modern sequencing technologies, massively parallel sequencing,
- it's now possible to read the entire human genome sequence in one sitting, in one experiment in a
- single day, and that's 3000 million base pairs of the human genome. So it's a really handy tool for
- cancer because human cancers are full of genetic changes or mutations, and if you could read that
- whole human cancer genome, you could learn a lot about what has driven each person's cancer.
- For as a cell turns from being a normal cell into a cancer cell, etched into the DNA of
- the cancers are thousands of mutations, the scars of historical events that have driven
- this cell into a malignancy. For the members of the audience that are not scientists, I'll try
- to visualise the extent of mutations in cancer. So here are the human chromosomes at conception,
- 22 pairs of autosomes and a pair of sex chromosomes. At the point of cancer diagnosis the
- chromosome complement can be so markedly evolved from its original state. So there's extra copies,
- there's lost copies, there are unusual new chromosomal combinations as well.
- Now, the fact that this scrambling of chromosomes, these chromosomal mutations
- were actually reported as far back as the turn of the century, turn of the last century.
- In 1902, Theodor Boveri first observed - using a good old microscope before we
- even understood about DNA - he observed in sea urchins that chromosomes were likely to
- be important units of heredity. They had to be present in the right quantities for
- normal cellular development, and he suggested at that time that cancers probably arose from
- cells that had had its chromosomes scrambled. Now this first notion was dismissed initially,
- and it took 13 years for others to verify Boveri's hypothesis. Had Boveri made the
- suggestion 100 years later, it would have been validated by someone else in 13 months in Nature
- Cell and Science, or if he was very good on social media he might have been able to deal
- with it through Twitter in 13 minutes, but 13 years for someone else to verify this idea.
- So chromosomes, although they provide us some insights into cancer mutations,
- they are rather crude, very low resolution, and you can unpack and unravel the chromosomes to
- reveal the DNA building blocks. That's adenine, cytosine, guanine and thymine, ACG and T.
- Just like the chromosome level mutations which we call copy number changes, you can have a
- panoply of different types of mutations. So substitutions are single-base changes.
- That's when you change one nucleotide for another. Insertions or deletions is when you remove some or
- you add additional nucleotides. Rearrangements are when you have a break in that double helix. That
- sugar phosphate backbone breaks off, relocates elsewhere and forms a new mutation. That's
- called a rearrangement. When we do whole genome sequencing you will be able to see them all,
- all these different mutation classes and all the mutations of those different classes.
- Now, for decades in cancer research, we have focused on identifying a small
- handful of mutations in each person's cancer - so-called driver mutations - because we believed
- these were causally implicated and drive the development of cancers. That's a fitting notion
- because these could be therapeutic targets. So here are a few examples of known driver
- mutations in cancer for which there are effective drugs that have been created, tested and are
- in clinical use today. Beyond the one to five driver mutations present in each person's cancer,
- there's thousands and thousands of passenger mutations, that which was not thought to be
- useful, just mutational noise, random events that are scattered throughout the genome.
- What whole genome sequencing technologies allows us to do is it gives us access. It allows us to
- see not just that small one to five mutations, but all the thousands of passenger mutations, and
- that's the sort of thing that I started doing in my research training at the Sanger Institute. Now
- I was set down a couple of times and I was sort of advised,
- 'Most people are studying driver mutations in the world today, Serena,
- and you're exploring passenger mutations. You're foraging around the bin of mutational detritus.'
- In fact, but in fact you can learn a lot by foraging around people's bins. You can learn
- who's eating too many pizzas and drinking too much alcohol. We're not here to judge. We just
- understand a bit more about how each person lives. So what I'd like to do in this section is to tell
- you a bit about how these tons of data that we get per patient from whole-genome sequencing, how we…
- It was only 21 cancers that we first sequenced, which doesn't sound like much, but at that time
- was a lot of data I'm going to share with you three major biological insights that we gained
- by just squeezing the information, the biological insights out of these first 21 whole genomes.
- So the first thing is about this DNA graffiti, mutation patterns,
- for as a cell turns from being a normal cell into a cancer cell there are these mutational
- processes that are constantly occurring. Some of them may be perfectly normal. As
- your cells divide you will acquire errors in DNA. As you sit here listening to me speak, your cells
- are exposed to oxygen and water. Those are the most essential elements of life, but they're also
- the most mutagenic. Oxidation and hydrolysis is happening in your DNA all the time. Now,
- there may be mutational effects from just being exposed to various things in the environment. So
- tobacco smoke is a good example of that, and we know that that can cause lung cancer. Ultraviolet
- radiation is also damaging and can give you skin cancer. There are DNA repair pathways in all your
- cells that are constantly mitigating this damage. Now, DNA repair pathways can go awry as well,
- and that can cause mutations. There may even be mutational processes with unknown causes for now,
- and we are constantly learning so we may start to uncover the aetiologies
- of some of those mutational processes. Whatever are the different mutagenic processes,
- each one leaves a characteristic imprint, a mark or a signature on the genome.
- When you sequence a cancer genome, what you see is a final portrait. It's a composite of all those
- signatures added together so it does look like a random collection of mutational detritus. In,
- fact when I was working in the lab of Mike Stratton at Sanger Institute,
- together with Ludmil Alexandrov, we showed that if you had multiple tumours of the same tumour
- type - in this case it was the 21 breast cancers - you could use mathematical methods to extract
- the pattern the mutational signatures - that's the pink triangles - and then you could even quantify
- the amount of each signature in each person's genome. That's determining the size of the arrows.
- Now, I'd like to show you some examples of what mutational signatures look like now.
- There's about 150 of these today so I don't have time to go through all of them. I just
- want to give you a flavour. We started with substitution,
- single base changes. So here are some examples of substitution signatures. We classically present
- them like this. There's six main substitutions, as you can see in the six different colours,
- and for every nucleotide your neighbours matter. So we take five prime ACGT, three prime ACGT,
- four by four is 16, 16 by six is 96. A signature is this 96-channel pattern, and
- for this particular signature where you see the tall bars, these are C to T mutations which occur
- at CPGs. So that's when a C precedes a G, guanine. These are methylated CPGs and they are constantly
- deaminating. This is a normal mutational process. It's happening in all your cells
- right now as you sit here listening to me speak. This is an intrinsic process, perfectly normal.
- There are some other intrinsic processes like these too caused by a family of enzymes called
- APOBECs. I hope you can see that they do produce quite distinct patterns. These C
- to T and C to G mutations from APOBECs tend to occur at a five prime T preceding the C.
- That's just how the APOBECs work. Now, here are some signatures that are due to environmental
- agents. The one at the top is due to UV light that occurs in skin cancers. The one in the middle is
- due to tobacco smoke, and you find those patterns in lung cancers, and the ones at the bottom is
- aristolochic acid, and that's seen when you ingest aristolochic acid. That's seen in liver and kidney
- cancers. So I hope you can see that these patterns are really quite distinctive and recognisable.
- That's a bit like street art, like graffiti art. Now, there are pieces of street art that nobody
- recognises, but there are some where you can know who the artist is. So this is Seen and
- he's been doing graffiti since 1970s, and he does comic-book street art. It's very distinctive and
- it's something you see a little bit in New York. This is Revok. He's got a very particular style,
- so recognisable unfortunately that he does have legal consequences to his art.
- Then this is Ben Eine. He's famous for his distinguishable alphabet letters that he's
- put all over shutters in East London. So he is celebrated for his street art. In 2010, British
- Prime Minister David Cameron gave Barack Obama one of Eine's canvases as a gift on a visit to the US.
- Just like in human cancers, mutations are there to be read. Here in this message… There is a message
- there. I don't know whether you see it. If you do, you can tell me about it later at the reception.
- So this is Banksy. Personal identity unknown, yet graffiti style very recognisable,
- iconic even. There is some political or social commentary behind it. Just like street art,
- with DNA graffiti you can tell whodunit. You can say from looking at the signatures. You can know
- what the aetiology is. Sometimes you can't. A bit like Banksy, we don't know his personal identity,
- but the significance of the of the signature in cancer can be delineated.
- So really the point I wanted to make there was that the power of data from
- whole-genome sequencing can help to reveal these mutation patterns as graffiti-like signatures,
- which may tell you something about the causes of each person's cancer.
- The second thing I wanted to highlight was that the graffiti-like patterns, they don't just have
- to be sort of evenly distributed through the genome. You can have some areas where there
- are localised mutations, hypermutations even. So again, let's visualise the data together.
- So let's say we lie our chromosomes down one to chromosome X, and number every single mutation
- from one to X, and we calculate something called an inter-mutation Distance. That's the distance
- from one mutation to the one immediately preceding it in the reference genome. Most cancers have
- about 3000 mutations, and the human genome, as I said is, 3000 million bases, so roughly you would
- have a mutation every million bases. That's roughly what you're seeing. You can plot this
- information like this. On the vertical axis we have the mutation distance, and that's on a log
- scale, and on the horizontal we have the mutation number. What you get is a cloud of dots like so,
- sitting at around 1 million base pair and these four dots there in that tiny little region right.
- You can colour the mutations like so to make it look a little prettier, but if you did have
- mutations that were clustered together, that were localised and there's focal hypermutation, you'll
- find much shorter inter-mutation distances like this. You might see it as a strip of mutations.
- Oh, yes, I have a pointer. You have a strip of mutations like so, and this sort of rainfall plot.
- Now, when we examined the 21 genomes we found this. This is a really dramatic example,
- but here are 800 mutations. They're all red dots. They're all C to T mutations.
- They're all in one place in the genome. In fact, there's another little strip here.
- So, this is something we call kataegis, localised hypermutation. Now the wonderful thing about
- whole-genome sequencing data is, it's digital. You can zoom out, but you can also zoom right in.
- So, those 800 mutations are on chromosome six, so we're going to look at chromosome six together.
- So, here's chromosome six lying on its side. That's a short arm and that's a long
- arm, so these are chromosomal coordinates, and this is your standard rainfall plot.
- Remember I said in whole genome sequencing you get all classes of mutations,
- so these are substitutions. Let's add on some other classes. To our surprise we find that
- it co-localises. The sort of storms or mutations co-localise with another class of mutation called
- rearrangements. So that's two different classes but they're all occurring in the same spot.
- Now let's look at this in higher magnification. We're going to go down 100-fold and 100-fold,
- one megabase, ten-kilobase windows, and now you're looking at individual DNA molecules.
- The yellow and blue rectangles are individual DNA molecules and the red bits are mutations. Now, I
- hope you can see that there are some DNA molecules where all the mutations are close together,
- and they're all on the same DNA molecule as well. This is giving us insights into how mutations
- arise. So if I draw it out for you as a schematic, by looking at whole
- genome sequencing data, you can see that these are C to T mutations. Interestingly,
- those C to T mutations are almost always preceded by a T thymine. You saw this before. I mentioned
- APOBECs. This is probably caused by APOBECs, and not only that. Those mutations are all happening
- on the same strand. Now APOBECs happen to require single-stranded DNA. APOBECs evolved to get rid of
- viruses. If it sees single-stranded DNA, it thinks it's virus is going to deaminate it. It's going
- to mutate it. So this gives us some really lovely insights into how mutations arise,
- some insights into mutational biology. That's really the second point I wanted to make,
- is that you can also learn quite a lot about new biology from this whole genome sequencing data.
- My last point is about that digital nature of whole-genome sequencing, because for every
- position in the genome you get many, many sequencing reads on each site. So you get
- digital information at each location. So, let's say you've inherited a mutation from your mum,
- or from your dad even. Here are your chromosomes, a pair of chromosomes for everyone. So, let's say
- you've inherited a mutation from your dad. If we zoom in to this site. Because one in two alleles
- is mutated, what you're going to see is 50 per cent of the sequencing reads are going to carry
- a mutation. That is if you have inherited it from one of your parents. In a cancer, you'll see a
- slightly different situation because when you take a cancer piece you will also capture some normal
- cells, some lymphocytes, some stromal tissue, some fat. So there will be some reads that are
- from normal cells, and let's say in this case 70 per cent of the reads are coming from the tumour.
- If you have acquired a mutation in your cancer on one of two alleles, half of 70 per cent is 35
- per cent. So, you know what to expect, four chromosomes that are present in two copies.
- So now you can plot this information here, and we know, for ,example in this patient
- chromosome ten is present in two copies. We expect to only see a cloud of mutations here,
- but I hope you can see that some additional clouds down here, additional mutations not present at
- where you expect it to be. This is evidence that in fact in your cancer, as it has developed,
- some new subpopulations are arising, new subclones are arising in the tumour, and those clouds of
- mutations down here is evidence that there's new mutations happening on chromosome ten in those
- little minor subpopulations. You can do that for all the chromosomes to infer phylogenetic trees,
- to try to understand the evolution of cancer. So this is a gross oversimplification
- for the reason of time, but basically you can use mathematics to model the mutations
- that you know have happened very early in cancer evolution. You can also model the
- mutations that have happened in the branches later. You can construct a phylogenetic tree,
- and that's from just one sample, if you have whole genome sequencing data.
- So these are the three main things I just wanted to communicate early on, which from whole genome
- sequencing data, in that first goal we were able to demonstrate mutational signatures,
- graffiti-like patterns. We can understand biology. We can explore cancer evolution. This
- I did in the labs of Mike Stratton and Peter Campbell, both fellows of the Royal Society.
- Since then, there's been a huge explosion in the field. So we're not the only ones
- doing this. Everybody's doing this now. What I just told you was when you get whole
- genomes from one sample, from one patient. People are now taking multiple samples per patient. You
- can draw up phylogenetic trees. That way you can take samples that are separated temporarily over
- time. So these are just different ways that you can draw up phylogenetic trees, and in the field
- of mutational signatures and DNA graffiti as well, from 21 genomes we then went to 500, and there
- was a landmark paper there, and then the field has just exploded. There's loads and loads of new data
- all the time. Now, I'd like to also point out that it's early days. We are still learning. Nothing
- is dogma. That knowledge and understanding will change because we started with very few samples,
- and as you have more data you'll get more knowledge as well. So there's no there's no pride
- in it. We'll be changing our views and changing our thoughts about certain things as we go along.
- There's also a lot of experimental work to support a lot of the mathematics. This isn't just
- mathematical hocus pocus. There actually is some evidence that you can reproduce these signatures,
- these graffiti-like patterns in a dish. I would also like to lastly point out that
- we could not have done this without the visionary and ambitious international sequencing endeavours
- that share their data. A lot of the work that we do is enabled by the fact that everyone shares
- their data. So I've sort of galloped through a first section of my talk. I'll give you a little
- bit of a breather with a tiny anecdote. Our work on kataegis made it on the cover of Cell, which
- was a highlight of my life, and I shared this with my mother. My mother is a Chinese lady who doesn't
- really care about Nature Cell and Science and she just wanted a daughter who was an NHS consultant.
- So, she was most unimpressed, but she did say… I was explaining mutational signatures and
- patterns to her and she goes, 'So you have a PhD in pretty patterns. You are not an NHS consultant.
- You are earning pittance. How does this benefit mankind?' Parents are good for grounding you,
- but this brings me on nicely to my second section, which is how we've now taken this data and try to
- create algorithms to try to get clinical value for patients. To do that, I'm going to walk you
- through more data, more visualisations of cancer whole-genome data. So I'm going to orientate you.
- Here's a patient. She's anonymised. We have her chromosomal ideogram, chromosomes one, two, three,
- four, five, all the way around to X and Y. Then we have the substitutions. Remember the drivers
- I mentioned? She's got one, p53, but beyond that one driver she's got 4500 other mutations with
- interesting patterns in them, including APOBEC. Now, insertions and deletions. We can summarise
- it like so. We can see that she has an excess of a particular kind of deletion which other
- people have shown to be associated with having a defect in a DNA repair pathway called homologous
- recombination, or HR. BRCA1 is a gene in the HR pathway, and this woman has inherited a
- BRCA1 mutation. So this fits. This pattern fits with the fact that she's got a BRCA1 deficiency.
- Going inwards some more, these are chromosomal copy number aberrations, what I started with. So,
- green means she's got gains in these chromosomes, and pink means losses. As you can see, she's got
- losses throughout the genome. Again, another pattern, another signature of BRCA deficiency.
- Now, rearrangements. She has over 300 of them. They are beautifully distributed around the
- genome. They all of a particular class called tandem duplications. This again is a signature,
- a pattern of BRCA1 deficiency. These are all patterns that are classic
- BRCA1 deficiency. They are the graffiti of BRCA1 deficiency, and sometimes referred to as BRCAness
- by the community. There are specific drugs that have been created for patients who have inherited
- mutations in BRCA1, that have BRCAness, drugs like platinums and PARP inhibitors in particular.
- So, if I showed you this patient again - I've just showed you her tumour, she's got
- inherited BRCA1 cancer. Hopefully, her tumour is going to be sensitive to PARP inhibitors - and
- I showed you these two other women, who are not related to her at all, and I said to you,
- 'The one in the middle has acquired BRCA1 mutation and the one at the end does not have any genetic
- defects. She has an epigenetic way of turning off her BRCA1 gene,' I think you'd agree that
- they might all be sensitive to PARP inhibitors. Although they don't share a single mutation in
- common - they each have tens of thousands of mutations, not a single one shared in
- common - but by taking this holistic cancer genome profiling approach you can instantly
- recognise the biological state of a cancer as being BRCA1-deficient, even if how they've become
- BRCA1 deficient is different. That's because what they do share in common are the graffiti-like
- mutational signatures. They are pathognomonic and distinguishing. This is the sort of information
- that is perfect for machine learning. Now I don't have time to go through lots and lots
- of different models. I just want to communicate the principle. When you have this sort of data,
- it's perfect for trying to teach models to be able to classify tumours for us so
- that we can enable the next generation to do cancer whole-genome interpretation.
- So I'm going to tell you about HRDetect, which is an algorithm that was designed to find BRCAness,
- because I've just told you that these tumours might be sensitive to PARP inhibitors. So the
- algorithm is HRDetect. We applied HRDetect on a cohort of 560 patients with breast cancer. Now
- to our surprise, we found a greater proportion of breast cancers that had high HRDetect scores
- indicative of BRCAness than we had expected. BRCAness wasn't limited to just one or five per
- cent of the cohort, which is what we expected. It was present in 22 per cent of the cohort.
- So, that's one in five breast cancers. Now, I know a lot of breast cancers are cured,
- but not all of them are cured. That's one in five breast cancers. This is a mix of breast
- cancers, the sort of thing that you get in the population. So that was a little bit surprising.
- So, we dived into it a little bit more and we found the 22 women that we knew had inherited
- BRCA1 and BRCA2 mutations, because we recruited them. So these are positive
- controls. We identified another 33 new families with germline mutations in BRCA1 and BRCA2,
- so opened a can of worms there. We found 22 patients who had acquired BRCA1 or BRCA2 defects,
- but one third of the cohort, we can see the graffiti-like patterns, we can see the signature,
- but we cannot find the genetic or epigenetic cause. So you cannot find the driver but you
- can see the patterns. The question is whether that is clinically valuable or not. So at this
- point we have an algorithm, an algorithm that has been developed through data science on the
- rich whole-genome sequencing datasets that are available in the community today, but to
- take an algorithm through to patients you need to demonstrate clinical value. We need to show that
- it can either prognosticate - that is to provide some sense of outcome - or better still, be
- predictive, that it can tell you whether someone's going to be sensitive to a particular drug.
- This is the hurdle. This is the slow step between discovery science and implementation
- in patients. This is the slow bit before getting to clinical utility.
- Nevertheless, we collaborated with colleagues in Sweden, because they have fantastic collections of
- clinical data, treatment and outcome data, and that permits us to try to answer the question
- of prognostication. So this was a new cohort of triple-negative breast cancers. This is a
- kind of breast cancer that tends to have a poor outcome. The project is called SCAN-B in Sweden,
- and it was led by Ake Borg and Johan Staaf from Lund University. Critically, they have clinical
- information, treatment and outcome data, and this was critical because it allowed us to
- show that HRDetect could prognosticate. It could distinguish patients who were going to do well
- from the ones who are going to do badly. This was irrespective of whether we could find a driver,
- a genetic or epigenetic driver. So, even in cases where you couldn't
- find a driver, the signatures, the graffiti-like patterns, were able to prognosticate. The next
- step in collaboration with colleagues in London - that's Nick Turner at the ICR - we explored
- our algorithm in a small proof-of-principle phase II clinical trial. We applied HRDetect
- in patients with newly diagnosed treatment-naive triple-negative breast cancer. They'd been given a
- window of PARP inhibition right at the start. Very unusual to get that sort of clinical trial. We
- did the genome sequencing at the start, and then Nick was collecting blood samples throughout. Now,
- if your tumour is sensitive to the PARP inhibitor, the circulating levels of tumour DNA is going to
- drop and that gives you a low CDR15 ratio. Sure enough, here you can see HRDetect
- high-scoring cases have got low CDR15 ratios. Now, this is a very small proof-of-principle
- study, but it allows us now to go towards a phase III randomised clinical trial.
- This shows us nicely that there might be early predictive capabilities of our algorithm as well.
- So I really use an example of one clinical algorithm to try to communicate that actually
- there are these additional steps before you can take it to patients,
- but it's important to do them, because otherwise they don't get to patients.
- Ninety per cent of algorithms don't get to patients. So this is the limiting step.
- Now, in recent years, The 100000 Genomes Project is an England-wide project that's sought to do
- whole genome sequencing in many rare diseases and in cancer. They've produced the largest cohort of
- whole-genome sequenced cancers in the world, of over 17000 NHS cancer genomes. Remember I started
- with 21? This is a huge number now, and together with all the other whole-genome sequenced cancers
- in the world, there's over 20000 of them. Here again I'd like to pause for reflection to
- make three points. I think we should acknowledge it's a spectacular feat by the NHS collective,
- that which we've been hearing such negative things about in the press, and I think it is pretty
- remarkable that if you invest in something for the medium to long term, you can mobilise a willing
- and able group of people. These NHS cancers were recruited from all over England and Wales,
- so you can make amazing things happen. I'm in awe that they did this project.
- On the backbone of that research project, the UK then led the way in implementing a national
- infrastructure called the Genomic Medicine Services, the GMS. Yes, it's early days,
- and beyond being an academic I am an NHS clinician. I am at the gritty end of the
- implementation spectrum, and I appreciate that there are issues, but beyond the operational
- issues I'd like you to see potential, the potential for a continuous cycle of learning,
- of using and reusing the ever-growing data, of making new discoveries or creating new
- algorithms and then performing the clinical studies and getting patient impact quickly.
- This is a system and resource like nowhere else in the world.
- The last thing I wanted to make, the last point I wanted to make at this point, was that if your
- cancer was used in any of these studies, the data are all anonymised. When I've showed you plots,
- they've all been anonymised. I don't need to know your name. I don't need to know where you live,
- but the data from your cancer genomes and the associated clinical information are incredibly
- powerful for academic learning. We can use it to turn them into tools and test them quickly, and
- we can bring benefits to patients faster. So, in our most recent endeavour mining the extraordinary
- treasure trove of nearly 20000 whole-genome sequences, including the NHS cancers, we're
- able to see such an incredible diversity of DNA graffiti or mutational signatures that are present
- in human cancers, and some of these can be used to provide very precise insights regarding how we
- might treat cancers more effectively per patient. With that, I'm going to come to my last section
- about how we gain new knowledge from large cohorts, but the idea is to bring
- benefits for individual patients. I hope you've seen that every cancer is highly
- individual. So we're going to walk through some real-patient whole-genome sequencing data.
- So here is a patient with a very typical breast cancer. You've seen these whole-genome sequencing
- plots. So beyond the driver mutations, the patient also has the signatures,
- the graffiti-like patterns, of BRCA2 deficiency. Now, if your clinician looking at whole-genome
- sequences is not very confident, apply one of the algorithms. What you get is a nice high score
- to help you in case you're uncertain. So this is great. We have got a BRCA2 deficient cancer. This
- tumour might be sensitive to PARP inhibitors. Great. Easy one. Here's a uterine cancer. We
- don't usually look for BRCA deficiency in uterine cancer, but here, look. It's got BRCA2 deficiency.
- It's got a p53 driver, BRCA2 signatures, a high score but no BRCA1 or BRCA2 mutations found.
- In today's clinic you would only be treating this patient for p53 mutation, which is basically no
- treatment at all. You wouldn't even see the BRCA2 deficiency. What could this patient get instead?
- What are we not giving patients, or what are we not exploring soon enough?
- Here's a colon cancer. This is an abnormality that's usually hunted down, which is mismatch
- repair deficiency. These are common in colon cancers, relatively common in colon cancers,
- and there's a driver mutation to go with it, an inherited MLH1 mutation. Now, we have many
- different algorithms. We've got another one called MMRDetect. So this patient has a high MMRDetect
- score, and then you can apply any other genomic assay, not just ours. Anybody's in the world.
- There's something called tumour mutational burden, TMB. This is an FDA-approved marker, a biomarker
- to identify patients with mismatch repair deficiency, because these tumours tend to
- be sensitive to immunotherapies. So that's great. That's an easy one. Here's a breast
- cancer again with mismatch repair deficiency, high MMRDetect score. It's got all the signatures,
- high TMB. This is not something we look for in breast cancer patients, so this patient
- will again not get necessarily the drugs that you might give patients with MMR deficiency
- right from the outset. You might discover it in due course, but these patients are basically
- treated as breast cancers. They're not treated as a breast cancer with a mismatch repair deficiency.
- That's really the message. Okay. Here is a uterine cancer with
- the graffiti-like patterns, the signatures of polymerase dysregulation. She has the acquired
- polymerase POLE mutation as well, and she's got a very high TMB. This is a classic polymerase
- dysregulated cancer. These cancers are sensitive to immune checkpoint inhibitors. That's an easy
- one. Here's an oral cancer. Oral cancers are on the rise in young adults in the UK. We don't
- really know the reason behind it, but it's also polymerase dysregulated and it would be missed.
- It's got a low TMB score, so it doesn't even fit that typical criteria. So again
- here might be a tumour that could be sensitive to immune checkpoint inhibitors that might just
- not be detected. Okay, last few. This is a lung cancer. I'm showing it mainly because the bluish
- tinge of tobacco smoke is very, very clear. Very clear graffiti sign of tobacco here.
- This is a kidney cancer of a patient who has had exposure to something called aristolochic
- acid. It's got the signatures. Interestingly, the patient doesn't think that they've had any
- exposure to aristolochic acid, so this is a public health thing. It's banned in the UK.
- This is a skin cancer, malignant melanoma with the signatures of ultraviolet radiation,
- ultraviolet damage. This is a very typical-looking malignant melanoma, the red tinge, high numbers of
- mutations, high TMB very clear. Keep that in your mind. If I showed you this tumour which
- was from a metastatic lung cancer, so there were lots of lesions in the lung, primary was unknown.
- If the primary is unknown and you're looking at this,
- wouldn't you think that this tumour might have come… The primary from the skin.
- Last but not least, I'm going to end on a clinical case. So I have the permission of the patient to
- tell this story. So, he's a 30-year-old male with a genetic inherited genetic abnormality called
- xeroderma pigmentosum, or XP. When you have an XP mutation you can't fix damage from UV very easily,
- so these patients tend to have photosensitivity. They have a lot of skin damage and they have an
- increased risk of skin cancer. He presented with a lesion on the medial aspect of his left eyebrow,
- and this turned out to be an angiosarcoma. That's a rare cancer of the blood vessel lining.
- So initially he had no metastases, wide local excision. You can see the skin grafts trying to
- heal the area. Unfortunately, over a relatively short period of time, he really didn't respond to
- any of the standard treatments for angiosarcoma. This is the usual furrow that we plough, isn't it?
- We have an angiosarcoma, we're going to treat it for an angiosarcoma. This is just the protocol.
- At this stage this man is really, really very ill. It's sort of terminal disease at this stage,
- and we do a whole genome sequence. You can see that tinge of red all around. That's
- ultraviolet radiation damage. You saw this earlier, but that's not a surprise because
- he's an XP mutant. You expect to see this. What's really killing this patient is the
- polymerase dysregulation that's present, but not in a very large amount. In a relatively small
- amount. We even find the acquired polymerase epsilon mutation that was caused by UV damage,
- but it's present in a very small amount of sequencing reads. Remember the phylogenetic
- trees we talked about. Present in a small variant allele fraction.
- This was present in a subclone of the tumour. So remember the phylogenetic trees I talked about.
- He's got his main tumour and then he's got his little subclone that has acquired this polymerase
- defect. Now these tumours tend to be sensitive to immune checkpoint therapies, and it was not
- easy trying to get this drug. It takes a village, and all the people at the bottom and who are in
- the audience today were involved in trying to get the patient the drug. We did all sorts of things,
- including staining of his primary sample, which did not show any evidence of antibody staining
- to PD-L1. There was a lot of to and fro in the discussions between us, and finally,
- PD-L1 staining was done in a metastatic sample because it's the mets that's killing him,
- and indeed that was all positive. So now we have really good evidence to support
- giving this man immune checkpoint therapy. I've made it sound very easy. This was a really
- hard battle. Here are his scans pre-treatment. He's got tumour everywhere. In his cranium, in
- his lymph nodes in his chest, in his liver. It was really everywhere and he was really sick. He was
- in hospital requiring pleurodesis. That's when you have to reinflate the lungs. Now, here are scans
- three cycles in after receiving pembrolizumab, and after seven cycles he walks home. In fact,
- today he's in the audience. He's alive and well. Thank you very much for being here.
- I'm sure he'll be happy to take any questions. So he exhibited a spectacular response. It's a
- great story, but really the message is this. In our sort of interactions with this patient,
- what he communicated was how he felt that he was treated as an individual with this cancer,
- as opposed to being another angiosarcoma on the standard sort of protocol.
- So with that, ladies and gentlemen, I have reached the end of my scientific narrative.
- All the work that you've heard about today cannot have been put together without the generosity of
- the patients, the families and all the healthcare professionals that collected all the samples which
- enabled our work. There's an extensive list of collaborators from really all over the world that
- have been unwavering in their support. There are so many I cannot list them all on the slide. It
- takes a village. It really, really does. There's the adage, 'Behind every man is a great woman,'
- but behind every woman is a critical group of mentors and sponsors that have looked out for me,
- that have put me forward for all sorts of things. These are great men and great women of all
- ethnicities that have supported, advised, pushed and challenged me over numerous
- coffees and Thai meals. Thank you for taking a chance on me. I don't think I was your typical
- person who would have gone into a PhD. I was in my 30s with two small children. I didn't
- have the most fantastic scientific pedigree, so thank you for giving me that opportunity.
- Our generous funders that have enabled my team and I to explore scientifically to our hearts'
- content. One or two of these funders go beyond just funding. You do put in huge
- efforts into growing and developing us. It's very hard to quantify how valuable that is,
- but it really is very valuable. Now, my team, past and present. A nicer bunch of
- people I cannot find. A refreshingly diverse, incredibly committed, thoughtful, sincere,
- secure and deeply interested group of people. Even when you disobey me and do loads of extra
- experiments and additional cell lines and way too many replicates, I am internally delighted.
- Your curiosity is what will drive innovation, and never let that be extinguished. I want you to know
- how much I am inspired by you. Thank you. Last but not least, the other team. The family and friends
- that are the sustenance of life. Your solid, grounded perspective I deeply respect. I feel
- so lucky to be able to bounce the daily grind off you. Packaged with your humour and zest for life,
- you provide the foundation from where I have grown. Now, my parents are no longer with us
- today. My mother was a was a very strong grounding force, and my father was the person who opened up
- my mind. I boarded a plane to the UK at the age of 17. I had never left Asia at that point. My mother
- said, 'It's all very well to get an education, but you must come home and marry a Malay boy.'
- I didn't do that. My father said, 'You know what? Whatever you hear behind you,
- whatever you hear from society, don't look back. The opportunities are ahead of you.
- They are over there. So grab those opportunities when they come, because that's the place. That's
- where you're going to learn new things. Don't look back.' With that, I'd like to thank the
- Royal Society again for this amazing recognition which has arisen out of all the opportunities that
- I've had in the UK. Thank you very much.
- utterly inspiring lecture. Fantastic science, straight into clinical medicine and the background
- of your group and your family. I have rarely heard such an inspiring lecture. It was wonderful. Thank
- you so much.
- So, now we have plenty of time. We have plenty of time for questions. So, ones online will come
- on to the iPad here, and as I see them I'll put them to our speaker. Let's start with
- some questions from the room. There's a roving microphone there, and there's a question here.
- Hi. Thank you so much for your inspiring talk. I was wondering, the dream of personalised
- treatment. How many decades do you think it will take the UK to get there?
- Well that's a huge question. Thank you very much for the question and for the kind words.
- You know, we already today have some evidence, today, in the audience. We have some evidence of
- it already in play, but it's too small. I think you're right. I think what your question is,
- when is it just going to become something that is part of the norm.
- I think we are unusual in that we have that infrastructure, that NHS Genomic Medicine Service
- infrastructure, where we can recruit patients, we can get information and we can send it back
- out very, very quickly. So, if you told me when I started this with 21 cancer genomes that we would
- have 20000 today, I would have been surprised. I would have thought, no, there's no chance we
- would have done that in a decade. I think we are a country that is unusual, and you
- have the infrastructure to be able to do this kind of personalised medicine pretty quickly.
- So I'd like to think that in ten years' time we'll be a lot further forward than we are.
- We have the infrastructure and you have very willing people. You have the GMS and Genomics
- England. I think it's possible to take it a good distance. We need to do the clinical studies.
- So just following up on that, how do the health economics stack up? How does it
- affect the cost of treating each patient?
- systematic studies on this probably need to be done. I think we need to look at… So of course,
- the cost. I think a lot of people were very concerned about the cost of genome sequencing
- initially. That was very, very expensive, and that actually has dropped dramatically. The limitation
- is the analysis and interpretation now, and then storage of the data. Because we have a
- centralised national infrastructure, storage of the data is becoming is less of an issue
- for sort of all the different sites in the UK. The analysis and interpretation is the tricky bit, and
- the costs there I think are something that that we need to understand. What we need to balance that
- with is, what is the cost for people when they are given months and months of the wrong treatment,
- or treatment that's not perfect? We give them rounds and rounds of certain chemotherapies. They
- are out of work. They're not able to contribute to the economy. We need to do those calculations.
- There's also the cost on the quality of life, of course, because if you give somebody a drug and
- they're really sick with it for a long period of time, how do you put dollars against that? These
- are all, I think, very important issues we need to address so that we can move the community,
- for I think there are still many people who think, 'Oh, this is still quite esoteric stuff.'
- I think we need to move people and shift that mindset, because we can read the whole genome.
- We just need to start doing it.
- Thank you. That was an absolute tour de force, and incredibly inspiring as an oncologist to listen
- to. As many of the drugs that we use in the clinic and their approval require a registration trial
- with a companion diagnostic, usually developed by a pharmaceutical company within that trial,
- can you say a little bit about how you could see this infrastructure and this whole genome
- sequencing triaging or informing how we move people, perhaps not to the immediate use of the
- treatment, but providing the triage into the trial that provides the evidence? That's the
- gap I think that's sometimes missing.
- what Professor Tutt is alluding to there is using this national infrastructure. Perhaps we can do
- whole-genome sequencing for every cancer. Triage. At that point you get the readout, and then you
- triage them into clinical trials so that you can get them through these registration trials faster.
- The UK is an unusual place. I think that is exactly the sort of thing that we can do,
- this sort of study. We'll need to talk to the regulators and also convince our colleagues,
- our clinical colleagues who will be doing all those clinical trials,
- that this is something that can be done. To do those, of course, there are the operational
- issues. We need to be able to make sure we can get samples which are not necessarily flash frozen,
- snap frozen. These are operational things. We can get results quickly. We can feed back
- meaningful information. So there's some operational issues for us to deal with,
- but I think that that notion that you've raised and something we have discussed,
- I think it's a fab idea. I think that's the way to try to get things through
- clinical validation studies and to patients as quickly as possible. There's nowhere else
- that we can do it. The UK is one of them. If you kind of think about mismatch repair deficiency
- and colorectal cancer, 45000 colorectal cancers in the UK, eight per cent are
- mismatch repair deficient. You'll be able to get thousands in a year and prove that you can give
- immune checkpoint therapy right at the start. So it's what we can do and we should do. Yeah.
- One here and then one there.
- Okay. So Serena, amazing talk, but not a surprise to me that you delivered that. My question is
- about prevention of cancer. So you focused on risk factors that have a major effect, which obviously
- is where you start with any investigation into disease, so tobacco, ultraviolet light.
- So how far away are we from being able to look at risk factors that have more modest effects,
- but which either over time or when they're additive may increase someone's risk of cancer.
- So, for example, environmental pollutants and obviously, in my area of interest, diet.
- So how far are we away from being able to pick up the effects such that we can prevent some
- of these cancers because we say, 'You know what? When you add that additive to food you
- increase the risk of this type of cancer'?
- Society, and we get Royal Society-level questions from members who are FRS. So, what a fantastic
- question. The information I've showed you today has been based on thousands of samples. To get
- that level of ability to be able to predict and to advise people and to intervene, and
- for something that has a smaller effect, you need a bigger population. You need bigger, many more
- samples. Then we need to do the data collection, because lots of these, especially at the start,
- were people's favourite samples in their freezer. To do the clinical studies I have to have a
- prospectively collected, right? I need to have the treatment and outcome data all done together.
- So what do we need to do to get to that ideal? We need to do a large study, and you don't have
- to worry about privacy because it's going to be anonymised hopefully, and you would collect
- environmental information, which is available. You can get Met Office information. You can know
- what the pollutant levels are. You can know how much pollen is in the air. You can know whether
- someone's going to get too much asthma in the population at any one time. So I think we are
- at an age where we should start to collect data and samples, and data from all sorts of things,
- not just the clinical data. The environmental data. Collect information about diet, behavioural
- scores. There have been one or two studies that have started to do this. The EPIC cohort,
- which is a European-wide thing, but also we've got an established EPIC cohort in Norfolk. I mean,
- that's a really good example of something that tried to do something like that,
- right, started off in that way. What's different today is that we know
- how much data we can collect, and it's a lot, and we need to structure that data so that we can get
- the next generation playing with that data, asking questions of it, finding the models,
- developing the models, and going, 'That's a high-risk population over there. They need to
- come for colonoscopy. That's a high risk. They need to have low-dose CT,' but it's a big study.
- I think the UK is one of those places that could do it, but we'll need to have the patients and
- the public on side because these are not people who are sick yet. These are healthy people, and
- so trying to motivate healthy people to come into a study is always a little bit harder. Hopefully,
- we should go out there and educate the public and try to get them involved.
- Outstanding stuff, Serena. Thank you for the talk.
- If money was no object, what would be next? What's your dream project?
- I do not have unlimited money with me, by the way.
- He's from one of the funders!
- What a great question.
- Where shall I start? I mean, I've got a massive list. I'll have to call you separately.
- So I think we've already heard some suggestions in the room today. One of that is more towards
- using genomics, or any kind of omics. I focused on genomes today, but actually you can use all sorts
- of omics data. So one bit is to really improve precision medicine, properly precision medicine,
- right. So getting those patients stratified right up front, sorting them out into clinical trials
- where we'll need approvals and people to be happy at NHRA level, regulators,
- etc, and then also having the pharmaceutical industry involved as well where they're happy
- to give out the drugs in this context, because what you heard from Professor Tutt there was,
- people do clinical studies in a very particular way. They have a very set
- mindset, and they also want to see the best results where patients are the sickest.
- So if a lot of breast cancers are cured, they don't really care. Not really. Not interested.
- We need to slightly change their mindset because yes, 70 per cent of breast cancers are cured,
- but they're not all cured. So, can we improve precision medicine for people with early disease?
- That prevention question was also a great one. If we really want to improve health outcomes, we know
- that we've got to hit… We've got to detect cancers as early as possible. So, I think that would be a
- nice space to try to explore. We've heard about the whole genomes. There's still plenty more we
- don't understand about the whole genomes and the germline. So when we have… If we're able to do
- a systematic study along the lines of what Professor Farooqi has just raised, I think
- that would be a really nice step forward. I think that'll be really enabling across the population,
- across the England and Wales-wide map.
- This is great.
- This one first one here, and then one at the front here.
- So the example you gave of the angiosarcoma, as I understand it from what you've explained to us
- before actually, the mutational signature had a lot in common with a malignant melanoma.
- Yes.
- are sequenced, is this a frequent thing, to end up with mutational pattern that suggests that it is
- similar to another one? Do you think that means that, for instance, a tumour like that actually
- did not actually arise from angioid tissue, even though it looks like it histologically,
- and actually what you've discovered is that it's actually just a very peculiar-looking melanoma?
- Treat the biology rather than…Yes, it may well have been a melanoma originally, but we should
- treat the biology rather than treat it based on tumour of origin. Now, you know we started
- with treating tumours based on tumour of origin, but that is perfectly reasonable for that time,
- right? That's the best we had and I think it's perfectly reasonable. If you have a breast cancer
- you're going to send them to a breast person. They're going to treat that appropriately. If
- you have an angiosarcoma you're going to treat that appropriately. If you have new information
- that now tells you, 'This angiosarcoma is not a typical angiosarcoma. It's got this
- abnormality instead,' I can tell you this is one of the biggest limitations. People will not come
- off the standard protocol. They will stick to that guideline because this is the guideline,
- and they're going to go through it, and then they're going to say, 'Okay, unresponsive to
- anything. Now let's look at the genome.' I think that's too late. I think we need
- to start to shift the mindset. We train our juniors to think like that,
- but I think in fact you raise almost two points, really. One is educating our next generation to
- think about how you adjust when you have new information, and this is quite new information,
- but also just educating the workforce in general. I think a lot… Genomics has now been slightly
- imposed on the NHS. We have to sort of implement it and we have to do it, but we haven't provided
- people with the tools to interpret, to analyse and to bring it back to patients, and we haven't
- provided them with the tools to deal with the results. So, I think there's a workforce thing we
- have to do to try to upskill and educate the next generation. I think you're really right. What you
- hit there is an issue that I've hit recurrently, almost weekly at the moment, which is the mindset
- shift that needs to happen to get people to come off that sort of standard protocol thinking.
- From a non-medical background, I'm very concerned. A family member, my brother, four years ago he had
- colorectal cancer. The initial treatment he had in a hospital in Leeds. He lives abroad
- in Andorra. So subsequently he's been monitored, and lesions, nodules, were found last year. Now
- should I possibly suggest genomic sequencing for him as a way to monitor potential cancers,
- or is that too extreme for me to interfere? He's at a very premier hospital in Barcelona,
- so I don't know if you can give me any advice that I can hand over to my brother. Thank you.
- So thank you for the question. I'm very sorry to hear about your brother. He does sound like he's
- in very good hands. Now, the vast majority of oncologists are doing fantastic jobs and
- if he's in Barcelona he's very likely to be in a very, very good environment as well. Genomics is
- offered in many places actually, and sometimes in colorectal cancers it
- is done. It's a little bit dependent on which country and which hospital. A lot
- of places offer it without even necessarily the patient knowing sometimes. In the NHS we do it.
- So it may be that that your brother and actually many NHS patients already get
- some level of genomics. They may not get this level of genomics, but they're already getting
- some level of genomics. I'm sure your brother is in very, very capable hands. If you wanted
- to have a separate conversation offline, I'm very happy to have that conversation.
- I think it's time for thanks and congratulations.
- Thank you.
- present you with the Francis Crick Medal, and to thank you warmly for the wonderful lecture
- that you've given us this evening, the food for thought and the vision of what treatment in the
- NHS for cancers might look like, hopefully in the near future. So congratulations.
Human cancers are highly individual. Etched into the DNA of cancers are graffiti-like mutation patterns, which could reveal underlying biological abnormalities, unique to each person’s cancer, with potential for application in precision medicine.
In this lecture, Professor Serena Nik-Zainal will describe how her team have explored the extraordinary DNA graffiti that has been seen in human cancers, using a combination of big data computational approaches and systematic experimental methods. She will provide an account of how they have designed algorithms that could be used to interpret cancer genomes for clinical purposes and how they have taken steps towards clinical validation studies for her algorithms. Professor Nik-Zainal will touch on her team’s recent endeavour, reporting the largest cohort of WGS cancers worldwide of nearly 20,000 patients recruited via the NHS. She will end by bringing the audience through a selection of real cancer WGS patient stories.
About the Royal Society
91TV is a Fellowship of many of the world's most eminent scientists and is the oldest scientific academy in continuous existence.
/
Subscribe to our YouTube channel for exciting science videos and live events.
Find us on:
Bluesky:
Facebook:
Instagram:
LinkedIn:
TikTok: