AI Makes Breakthrough in Cancer Detection
#ai #artificialintelligence #breakthrough #scientists #research #cancer #cancertreatment
Google's artificial intelligence has just discovered a combination of anti-cancer drugs that has never been tried before in the history of medicine—and proven that it actually works on living human cells.
Currently, millions of patients don't respond to immunotherapy because their tumors remain "cold"—that is, invisible to the immune system. The body could destroy the cancer itself, but it simply doesn't "see" it.
Scientists have long known that a protein called "interferon" helps make tumors visible to the immune system. The problem is that most patients have too low interferon levels for this to work.
That's why Google and Yale University created a massive neural network, C2S Scale, with 27 billion parameters. It was tasked with finding a drug that boosts the interferon signal in patients with low levels.
The model had to think "hypothetically"—that is, find a drug that enhances the effects of interferon, but that does not pose any danger on its own.
The AI virtually tested over 4,000 drugs in two scenarios: on real tumors from patients with a weakened immune response and on isolated cancer cells in the lab.
Ultimately, the model found the drug Silmatacertib. This is not a new drug—doctors have already used it for other diseases—but the AI predicted that, when combined with a low dose of interferon, it would significantly increase the tumor's "visibility" to the immune system.
Previously, no one in the history of medicine had linked this drug to the immune response. In other words, the AI didn't simply replicate what was already known; it created a fundamentally new idea.
To verify the model's accuracy, the researchers tested the results in the lab on living human cancer cells. And when silmatasertib was combined with a low dose of interferon, tumor visibility increased by 50%—exactly as predicted by the artificial intelligence.
The model has already been made publicly available, and this is one of the first cases of AI making a real scientific discovery that human logic hadn't anticipated.