Google has launched a new AI model named Cell2Sentence-Scale 27B (C2S-Scale) in partnership with Yale, which could pave the way for new cancer therapies.

The team tested 4,000 drugs using patient samples, allowing the model to predict which ones would effectively enhance antigen presentation in a relevant patient context.

The AI model managed to identify 10-30% of drugs that were already known, while the rest were completely new.

C2S-Scale 27B Model

This model, built on the Gemma framework, boasts 27 billion parameters and successfully generated a fresh hypothesis about cancer cell behavior, which the research team later validated as accurate.

The AI model is crafted to comprehend the “language” of individual cells.

One major hurdle in cancer immunotherapy is that many tumors are “cold,” meaning they go unnoticed by the immune system.

To make these tumors detectable, they need to be prompted to show immune-triggering signals through a method known as antigen presentation.

The C2S-Scale 27B model was tasked with identifying a drug that could serve as a ‘conditional amplifier,’ enhancing the immune signal in a specific “immune-context-positive” environment where certain key immune-signaling proteins were already present.

The smaller AI model lacked the reasoning skills to tackle this issue, a larger version of the model could eventually succeed.

Google has made the C2S-Scale 27B model accessible on GitHub and Hugging Face.

Subscribe My Channel





Discover more from Connect2ConnectOnline

Subscribe now to keep reading and get access to the full archive.

Continue reading