You do not get cancer from one error in your DNA. Only critical combinations of collaborative DNA changes (or mutations) can disrupt a healthy cell in such a way that it starts to behave like a tumor cell. We know that these mutations have to work together, but exactly how they do this is largely unknown.
To answer that question, researchers from the Netherlands Cancer Institute and the University of California San Diego School of Medicine used large-scale statistical algorithms to analyze the DNA of patients with cancer, from which patterns emerge of mutations that occur together in a cancer cell surprisingly often, or of mutations that are rarely seen together. This exposes the important partnerships of mutations that are underlying to cancer.
Their study, published May 30, 2019 by Cell, demonstrates that the patterns previously assigned to direct relationships between mutations within the cell can almost always be explained in other, indirect ways.
By taking these alternative explanations into account, in the future researchers could better recognize which combinations of mutations are important for cancer development and which are not.
"Now that we know this, we can develop algorithms that take this into account, which will greatly improve our understanding of cancer," says doctor-researcher Joris van de Haar, MD, from the Netherlands Cancer Institute, who co-conducted the research at both institutes.
A lot of research is being done worldwide to unravel the underlying rules that determine which combinations of mutations lead to cancer and which do not. Since millions of different mutations are possible, and not all of their functions are known, this is a hugely complex problem.
Joris van de Haar: "Almost all of the knowledge we have in this area has been found by studying cell lines and mouse models. In this type of research, for example, certain combinations of mutations are made in the DNA of a mouse and then examined to determine whether this leads to cancer. Although this has taught us a great deal, we all know that it has limitations. After all, mice are not real patients and the rules can be different. It is also impossible to provide mice with every conceivable combination of mutations."
This has caused a second branch of study to emerge, involving large-scale statistical analyses on DNA data from cancer patients. We look at which combinations occur surprisingly often: these are then the combinations that together would be necessary for the development of cancer. This can be compared to buying the ingredients for a meal: rice, coconut milk and curry paste are often bought together because you need them all for the same dish.
Combinations that are rarely seen together are just as interesting because they can be mutations with the same functions. If the tumor already has one of these mutations, it no longer needs the other. Just like you usually do not need two different brands of coconut milk at the same time. In this functionalistic way, you can categorize the millions of mutations that should have comparable functions.
"This sounds very logical, but if we critically evaluate the results of this approach, it has only very marginally improved our understanding of cancer so far," says Joris van de Haar. "We are now exposing important factors as to why that is."
The researchers show that there are almost always other, indirect factors that determine which mutations often occur together. Most previous findings had nothing to do with the relationships between mutations. Van de Haar: 'This problem is so complex that it has misled the smartest researchers for years. Now that we are aware of the problem, we can develop algorithms that actually do teach us how the partnerships between cancer mutations work."
Since combinations of DNA mutations are the cause of cancer, they are the key to understanding and treating this disease. It is, therefore, becoming more common for cancer patients to receive "personalized" treatment with drugs that specifically target the mutations in their tumor cells.
Van de Haar: "The current strategy of personalized medicine is rather simple: with this type of tumor with mutation A, we give medicine X. That may be effective, but it ignores the complexity of this disease, because a tumor also has a unique set of other mutations, in addition to mutation A, that you ignore. Ultimately, you would like to understand how all those mutations interact so that you can focus treatment on the complete mutation profile of a patient. Although this is still only at the horizon, we are one step closer."