There is a lot of controversy with «mundane» artificial intelligence, but when applied to medicine, it can be a great help to humanity. It is the case of Cell2Sentence-Scale 27B (C2S-Scale 27B), a Gemma-based AI created by Google DeepMind and Yale University, which has made a historic discovery for cancer treatment.
C2S-Scale 27B has proposed a method that makes it possible to convert «cold» tumors, which are difficult to detect, into «hot» tumors that are more attackable with anti-cancer therapies.
The method has been clinically validated, and will allow creating new medications and treatments to attack different types of tumors that until now were only detected when it was too late.
According to the researchers, it is a very important discovery, because it explains a behavior of cancer cells, in their interaction with antibodies, that was not known until now.
Google’s AI that helps cure cancer
C2S-Scale 27B is a 27 billion parameter artificial intelligence created by Google DeepMind and Yale University. She is based on Gemma, Gemini’s «mother», but has been trained with medical data to predict the behavior of a single cell.
As Google explains in the press release, one of the main challenges of cancer immunotherapy is that many tumors are «cold», that is, invisible to the body’s immune system.
A key strategy to make them «hot» is to force them to emit signals that activate the immune system through a process called antigen presentation.
«We asked C2S-Scale 27B to find a drug that would act as a conditional amplifier, one that would enhance the immune signal only in a specific environment, in which low levels of interferon (a key protein for immune signaling) were already present, but insufficient to induce antigen presentation on their own,» explains Google.
This task requires a level of conditional reasoning that smaller AI models cannot solve. But yes the 27 billion C2S-Scale 27B parameters.
The researchers designed a «virtual dual-context display,» which simulates the effects of more than 4,000 drugs on tumor samples from real patients.
A practical discovery of AI in the fight against cancer
The AI was able to create a drug capable of enhancing the body’s immune signal in specific contexts.
Not only is it an important discovery in itself. It is very relevant that C2S-Scale 27B proposed a method that was not known until now. Clinical tests have shown that it works. It will be very useful in the creation of therapies capable of attacking «cold» tumors, which until now were not detected until it was too late.
Honoring how science works, the C2S-Scale 27B AI model, which fights canceris available for download on Hugging Face, completely free and open source, so that any medical team can use it.
Get to know how we work in NoticiasVE.
Tags: Artificial intelligence, Health, Industry