Google’s AI C2S-Scale:
Google DeepMind’s AI model Cell2Sentence-Scale 27B (C2S-Scale) produced a new, lab-confirmed hypothesis on cancer cell behavior, marking a breakthrough in AI-driven drug discovery and biological research.
- C2S-Scale is a large language model (LLM) built on Google’s Gemma-2 architecture, trained to understand gene expression as a language.
- With 27 billion parameters, the model can capture subtle relationships among genes, cells, and tissues, reflecting AI “scaling laws”—where larger models gain emergent abilities that smaller models lack.
- It translates single-cell RNA sequencing (scRNA-seq) data and interprets biological functions as ‘cell sentences’, learning patterns from millions of cells to understand cellular functions.
- The AI hypothesized that the drug silmitasertib could act as a conditional amplifier, making cancer cells more visible to the immune system only in the presence of low interferon.
- C2S-Scale enables in-silico (computer-based) screening at unmatched speed and scale, accelerating scientific discovery.
