The algorithm that beat prostate cancer at chess

Three years ago, an advanced artificial intelligence managed to beat the world champion in the game "Go", using strategies and steps that humans had not thought of before - even after more than two thousand years of historical experience in the game. And the question everyone asked was - so what?

And really, so what? So what if artificial intelligence manages to beat humans in a board game?

We are beginning to see the answer today In a new study, which showed that sophisticated algorithms are able to manage strategies that beat cancerous tumors at their own game inside the body. The results were so impressive that the researchers are currently starting wider tests, and plan to apply the algorithmic strategy to other types of cancer as well.

In order to understand the role of the algorithms in this matter, one must first understand what causes cancer patients to die in the first place. The original tumor is not the one that causes death, usually. But at a sufficiently advanced stage, the same tumor overgrows cells that move into the bloodstream, settle in tissues throughout the body, and begin to grow secondary tumors that develop rapidly and kill the patient.

The researchers - Robert Gatenby and his colleagues from the Moffitt Cancer Research Center in Florida - developed an algorithm that relies on information from the clinic and treats the whole matter of cancer in the body as... a game. just playing And the rules are simple: the enemies are the cancer cells and the goal is to stop their growth in the body.

In the early days of the war on cancer, when scientists were just beginning to understand the disease, the answer seemed clear and simple: we must kill the cancer cells with the most powerful tools we have. And so drugs were developed that are supposed to eliminate the cancer cells. But a problem soon became apparent: these drugs killed a large part of the cancer cells, but the remaining ones - the remainder of the discharge - underwent a rapid evolution and developed resistance to the drug. Then, taking advantage of all the extra space that was freed up when their weaker friends died, the resistant cancer cells were able to continue to grow and spread through the body unhindered. At this point, the doctors would admit defeat, throw up their hands and tell the patient that his time is up.

We now understand better the mechanisms of cancer evolution, and the ability of cells to acquire resistance to drugs, so oncologists are developing more complex ways of administering anti-cancer drugs. They try to do this by relying mainly on the limited human intelligence. But what if we could harness advanced computing capabilities to develop strategies that would win the game against cancer - or at least lead to a draw?

To test the idea, oncologists conducted a study on prostate cancer patients. They measured the growth of cancer cells every month by tracking the substances the tumors released into the bloodstream. A dedicated algorithm calculated the dose of a certain drug - abiraterone - to be given to each patient, according to the rate of spread of the tumor. This drug kills the cancer cells that produce testosterone - a substance that the prostate tumor needs like air to breathe in order to continue growing. The trouble is, too much abiraterone will kill all the testosterone-producing cells, and push the tumor to undergo evolution: new cells will emerge that don't need testosterone. They will be stronger and freer to spread in the body - and they will also be the ones who, in the end, kill the patient.

The greatness of the algorithm is that it found the exact dose of Arbitron that should be given to the patients every month, so that the testosterone-producing cells suffer - and with them the entire tumor - but not all of them die. And because of this, the tumor remains relatively small, and does not develop the more deadly cancer cells that would kill the body. The new strategy worked impressively on the 17 patients, doubling the amount of time required to develop resistance to the usual drug dose, and doing so while using only half the amount of drug normally required.

There is, of course, criticism of the study. Despite the fact that it was published in the respected scientific journal Nature (Nature Communications), it is still only a preliminary study conducted on a very small sample of patients. It is important to test the ideas behind it on a larger number of patients. It is also important to clarify that there is no sophisticated artificial intelligence here that developed a new strategy, but rather that humans thought of the idea of ​​adjusting the treatment - and the algorithms were only able to adjust the treatment to each patient easily and efficiently.

But this is just the beginning.

More sophisticated algorithms will be invented in the near future, and will lead to the optimization of the drug regimen given to cancer patients (and in general). Even more advanced artificial intelligences will develop complex and efficient strategies that will play chess, checkers and go with cancer cells and HIV viruses, with Alzheimer's and Parkinson's diseases - and even with the body's aging processes. The rapidly developing digital technologies will help us use existing medicines in new ways that will result in much more effective results. Instead of focusing solely on the development of new drugs - a complex and expensive process that requires over a decade for successful implementation - new and smarter treatment regimens will be developed, driven by the power of artificial intelligence.

Artificial intelligences of this type, which we hear about in the research news more and more often, will help save the lives of many in the coming decades. And then, in the future in the long run, we will also fight them in the biggest game, in the face of death itself. We will challenge them to find the treatments that will extend the healthy life span of humans, and stop the aging process. Admittedly, there are still many decades until we will be able to sufficiently understand the aging processes of the body, or until we will develop precise enough tools to influence it, but we have already started walking on the road there, and if some catastrophe does not occur for all of humanity, we will surely also reach the desired destination.

for life!

You are invited to read more about the future of medicine and artificial intelligence in my book "who control the future", in the selected bookstores (and those that are just fine).