Google DeepMind has just announced an important scientific advance in the field of medicine, an artificial intelligence (AI) model that can decoding the working mechanism of cancer cells.
In collaboration with Yale University, the research team developed C2S-Scale 27B, a 27 billion-parameter platform built on the open Gemma architecture, with the goal of understanding the language of each cell.
According to DeepMind, this is the first time an AI model has not only analyzed biological data but also proposed a new scientific theory on the behavior of cancer cells, which was later confirmed to be experienced on raw cells.
In other words, this AI does not simply learn from data, but it can reason to find new directions in cancer treatment.
C2S-Scale 27B has predicted potential drugs that can trigger an immune response against tumors, by simulating how cancer cells interact with the immune system at the clinical level.
After being tested on human cells, some initial results showed the ability to open up more accurate and personalized immunotherapy in the future.
One of the major challenges of immunotherapy today is that tumors often "hide" from the immune system, making it impossible for the body to recognize and destroy them.
The goal of the model is to identify drugs that help amplify immune signals, forcing cancer cells to express immune-activating signals through the process of presenting antibodies. This is a key factor for the immune system to recognize and attack tumors.
To do this, the research team used a method called "dual-concept virtual screen", allowing AI models to analyze more than 4,000 drugs on thousands of different tumor samples.
Through millions of simulations, AI can identify compounds that enhance immune activation under specific cell conditions.
Notably, among the drugs proposed by the model, some are completely new, unknown, related to immunotherapy.
A typical example is CX-4945, which is predicted to help increase the risk of developing ancient insects by more than 50%, opening up new treatment prospects for many types of cancer.
Scientists say that with the ability to simulate and reason at the cell level, Google DeepMind's model could change the way drugs are researched and developed, from having to conduct expensive real-life testing to AI-based virtual laboratory simulations.
This discovery not only strengthens the role of AI in life sciences, but also shows how technology can help people get closer to completely decoding cancer, one of the biggest medical mysteries of the century.
