Google DeepMind and Yale University have unveiled a groundbreaking artificial intelligence system that could accelerate cancer treatment research. The C2S-Scale 27B model identifies complex biological patterns within cells and predicts how various drugs will affect tumors.
This AI has uncovered a mechanism that renders "cold" tumors visible to the immune system. Typically, such tumors evade detection by the immune response, complicating treatments, including immunotherapy. This new discovery may lead to previously impossible combination therapies.
C2S-Scale 27B analyzed over 4000 drugs on patient tumor samples and in laboratory settings. The model was able to identify substances that selectively enhance the immune response without affecting all cells indiscriminately.
Among the most intriguing findings is the CK2 kinase inhibitor silmitasertib (CX-4945). The AI predicted that the drug is effective only in the presence of a small amount of interferon in the cells. Both the drug and interferon alone had minimal impact on the cells, while their combination boosted immune system activity by 50%, "warming up" tumors and making them visible to the body's protective cells.
The AI's predictions were validated in the lab using human neuroendocrine cells that the model had not previously encountered. This demonstrates that the system can not only process data but also draw conclusions based on the context of the cells.
According to scientists, large AI models could serve as "virtual laboratories" conducting thousands of simulations and uncovering unknown connections between drugs, cells, and the immune system.
"This discovery could pave a new path for cancer treatment development," said Google CEO Sundar Pichai.
The success of C2S-Scale 27B illustrates that AI can accelerate scientific research and assist in creating effective treatment methods much faster than traditional approaches.
Recently, researchers from New York University Abu Dhabi (NYUAD) also introduced a new type of brain implant capable of delivering drugs simultaneously to multiple brain areas with high precision.