The C2S-Scale 27B model, one of the most advanced and scalable AI systems, is based on the Gemma model family from Google and is aimed at studying the interaction between cancer cells and the immune system.
In a large-scale study, it was found that vitamin D can slow down aging processes by three times. The model has the ability to "read the language" of cells and identify ways to make hard-to-diagnose tumors visible to the immune system. The problem is that such "cold" tumors often go unnoticed by immune cells, which is one of the main challenges in oncology. The development proposed by DeepMind opens new horizons for treatment, offering ways to "heat up" such tumors.
C2S-Scale 27B conducted an analysis of over 4000 drugs using a two-context virtual screening method. The model's task was to identify substances capable of enhancing immune signaling under specific biological conditions. About 10-30% of the compounds found became new candidates for immunotherapy, including silmitasertib (CX-4945), a CK2 kinase inhibitor.
Testing the hypothesis on human neuroendocrine cells confirmed the predictions made by the AI. In combination with low doses of interferon, silmitasertib increased antigen presentation by 50%, allowing for the activation of immune recognition of tumors where it was previously impossible. Alone, these drugs did not demonstrate significant effects.
This discovery emphasizes that C2S-Scale 27B analyzes not just data, but their context, revealing how different cellular states affect treatment efficacy. This approach opens new strategies for combating tumors that are resistant to existing immunotherapy methods and creates a virtual laboratory for further research. Scientists continue to study the mechanisms identified by the AI and test other predictions of the system.
