The raw material with which the biologist Núria López-Bigas (Monistrol de Montserrat, Barcelona, 47 years old) works is data. Thousands and thousands of data. In her laboratory, the scientist, who is an ICREA research professor at the Barcelona Biomedical Research Institute and an expert in bioinformatics, genetically sequences real tumors and dissects that information through powerful artificial intelligence (AI) tools. In these huge amounts of data may be the answer to why a cell goes out of control to create a cancer or what gene mutations are precursors of a tumor. “We help to digest all this information efficiently to give a report that is useful to clinicians”, sums up the researcher, who was recently awarded the Lilly Foundation Award for Preclinical Biomedical Research 2023.
López-Bigas has many open fronts. The first, deciphering which mutations, of the thousands that occur in cells every day, are drivers (lead to cancer) and reveal the genes that can cause tumors. But he also wants to shed light on the early stages of cancer development, what happens molecularly so that someone healthy ends up suffering from the disease later on. For one thing and the other, López-Bigas needs data and looks there for genetic patterns, warning signs that answer all those questions. He has already analyzed the sequence of some 33,000 tumors of 70 different types of cancer and has a list of 600 suspected genes: “We know that when these genes have mutations, they can cause one cancer or another,” she explains. But much remains to be known, admits the scientist, who dreams of getting to analyze a million tumors.
Ask. These are bad times for artificial intelligence, but you are on the good side of history. How can artificial intelligence help cancer?
Answer. Very much. AI is nothing more than a set of tools to extract information from a large amount of data. In our case, the large amount of data is tumor genomes and clinical information from these patients. And the AI are algorithms that allow us to extract a series of patterns, make automatic classifications, etc. Artificial intelligence by itself is not bad, but it depends on how you use it.
Q. How far can you go?
R. It has enormous potential. A genome is very large: when we can take the genome of a tumor and sequence it in its entirety, we find thousands of mutations. But of these, only a few are the cause of cancer. We are interested in finding these mutations in the midst of the thousands that this genome has and this is very difficult; it’s easier if, instead of having one genome of a tumor, we have thousands of genomes and we start to see a series of patterns that tell us where those mutations are. In the latest work, what we have done is, now that we know that this gene is important for developing cancer, we know which specific mutations in this gene cause cancer and with which mutations, even if it has them, nothing happens to it.
Q. What is the use of detecting these mutations? drivers?
R. When we know what mutations are drivers in a tumor or in a specific gene, it is giving us information about the molecular bases of cancer, what does not work in that cell for it to behave in this way. This gives you ideas of how to counteract it: for example, some of these mutated proteins are therapeutic targets. Other information that it gives is in personalized medicine: when we already have some therapies that can be directed against this protein, we are interested in knowing if this specific patient can benefit from it and you have to look to see if there are mutations in that gene that could be a biomarker for that specific treatment.
Every day mutations can accumulate in our cells by chance and for thousands of reasons.”
Q. Two breast tumors are not the same. Because? What happens at the molecular level so that, being in the same place, they are different?
R. It is one thing for it to be a tumor that starts in the same type of tissue and another thing is, at the molecular level, what has caused this cell to become a tumor cell. And part of this has to do with what mutations it has so that some proteins don’t work or work differently and this cell behaves differently. In the same tissue, it can happen that (a cell) has been transformed by some alterations and in the other tumor, by others.
Q. Mutations always have a point of chance.
R. Every day mutations can accumulate in our cells by chance and for thousands of reasons. If we are exposed to sunlight, this damages the DNA of skin cells and this can generate mutations. But because the genome is also so large, there are thousands of mutations, but only a few lead to cancer. There are many cells in our body with thousands of mutations that behave completely normal.
Q. What is special about those mutations that are precursors of a tumor?
R. They are mutations that affect key genes in the function of division or proliferation. The genome is very large and only 2% is made up of protein-coding genes. So, if there is a mutation that falls in 98% of the genome, nothing happens anyway.
Q. How is it possible for a cell to jump over all the control barriers that the organism has, go crazy and start reproducing uncontrollably?
R. This is like a continuous game of variation and selection, which are the bases of Darwinian evolution. The variation is given because there are mutations that are generated randomly: each cell in a colon epithelium, for example, can be slightly different. If a cell acquires a random mutation in a gene that still makes this cell divide faster, there will be more of these in this epithelium. After a while, perhaps one of these cells that already has this mutation, acquires another one that allows it to not only divide a little faster, but also to invade some tissue. It’s like a continuous process of variation and selection until the moment a cancer is diagnosed. When that happens, these cells have a long history. How have the barriers been broken? It is a process that does not happen in a day, it is more gradual. One question we might ask ourselves is why we don’t have cancer every day. Because there are all these barriers.
Q. Are attempts at tumors that are stopping being generated all the time?
R. Exact. Mutations are continually being generated, the immune system is continually detecting cells that behave abnormally, that have mutated proteins, and eliminates them. We don’t have cancer continuously because we have a series of barriers, including the immune system, to avoid it.
Q. If there are mutations all the time, are the mutations drivers are they infinite?
R. One thing that has been seen recently is that healthy tissue also has mutations and also has mutations that lead to cancer. With age we have more mutations because they are accumulating every day and if we now take any person without any tumor and look at the skin or the esophagus, we find mutations that could cause cancer, but those cells look totally normal. One of the big questions we now have is how these cells are still behaving normally and what causes them to eventually transform and generate a tumor.
Q. By analyzing the genome of the tumor, do they end up discovering the life that someone has led?
R. When we sequence the genome of a cell or a tumor, we are looking at the entire history of this cell’s lineage: the mutations it has accumulated over time, what things it has been exposed to… For example, we could tell, in some cases, if it has received chemotherapy or if they are cells that have been exposed to the sun or to carcinogens, such as tobacco, by the pattern of mutations.
Q. But now not all tumors are sequenced.
R. No. Now more is done when it is considered that there will be an added value to decide one treatment or another. It is not done in all types of tumors or in all people because, in some cases, right now the best treatment is the one you already know.
If we now take someone without any tumor and look at the skin or the esophagus, we see mutations that could cause cancer.”
Q. If the molecular profile of the tumor is so important and not all of them are sequenced, are you going a little blind in some cases?
R. We are still at the beginning of everything that can come from the application of personalized cancer medicine. I think if we start to imagine one day where the whole genome of tumors will probably be sequenced more routinely. But to get here we need sequencing costs to drop, for us to know how to better interpret this amount of information and for there to be a transformation in the clinic to see the value of that, understand it and know how to use all that information. It is a matter of time before all this happens.
Q. Earlier you said you were looking at DNA that codes for proteins, but what about that 98% of so-called junk DNA? To what extent can you modulate everything else?
R. There is very important information in sequencing all this extra 2%. We have analyzed entire genomes of 7,500 tumors and then up to 33,000 exomes, which is 2%. In the clinic, less than 2% is done, only 60 or 100 genes are sequenced, which are the ones that currently have therapies against them and are mutated. What we see is that there is a lot of information that we can extract when we have the entire genome. For example, we know things about mutational patterns and we see them much more clearly if you have the complete genome. And some of these patterns tell you if you’ve had chemotherapy, what is the number of mutations accumulated by that treatment… It’s a much broader read, but it’s a bit overwhelming, especially in the clinic, having all this information if you don’t know what to do with it.
Q. Are they capable of interpreting what that 100% of the tumor genome says?
R. Every day we know how to interpret more things. It’s also like a fish that bites its tail: the more data we have, the more information we can extract that helps us better interpret the next patient.
Q. Where is the knowledge hole now?
R. For the clinic, the most relevant question is what is in this genome information that can inform us of the weak points of that tumor. The other is to have information on how the tumor is going to progress, as if we could look into the future.
The immune system is continually detecting abnormally behaving cells and eliminating them.”
Q. How is the molecular image that you have now of the tumors?
R. It’s not very sharp. In the last article we showed that, using this AI, we can build models that allow us to distinguish between mutations. drivers or passengers. And we’re presenting this as a proof of concept that if we have data and we build the right algorithms to learn, we can learn this. But we need a lot more data.
Q. Can AI help to know if a person is going to develop cancer?
R. No, not this yet. The application that it could one day have is more like prevention, not to identify a person who is going to have it, if not, if we understand a little better what generates this risk and how healthy tissue is modified when it is exposed to it, we can still better understand how to prevent it.
You can follow THE COUNTRY Health and Well-being in Facebook, Twitter and instagram.
