Google’s AI Makes Cancer Drug Discovery Breakthrough

Google’s new AI model has successfully identified a novel cancer therapy approach that has been validated in laboratory tests, marking a significant milestone in AI-driven scientific discovery.

Key Takeaways

• Google’s C2S-Scale AI model discovered a new potential use for existing drug silmitasertib in cancer treatment
• The AI-generated hypothesis was experimentally confirmed in living cells
• This approach could dramatically shorten drug discovery timelines

Revolutionizing Drug Discovery with AI

Google’s Cell2Sentence-Scale 27B (C2S-Scale), built on the Gemma open models, represents a 27-billion-parameter foundation model designed to understand cellular language. The AI suggested using silmitasertib to enhance the immune system’s ability to detect early-stage cancerous tumors.

“This announcement marks a milestone for AI in science,” said Shekoofeh Azizi and Brian Perozzi, Google DeepMind and Google Research scientists.

“C2S-Scale generated a novel hypothesis about cancer cellular behaviour and we have since confirmed its prediction with experimental validation in living cells. This discovery reveals a promising new pathway for developing therapies to fight cancer.”

Novel Application of Existing Drug

While silmitasertib is already in clinical trials for multiple cancers and has orphan drug status for cholangiocarcinoma, Google’s breakthrough lies in the AI’s ability to scan vast cancer literature and identify new therapeutic applications.

Pharmaceutical companies typically spend billions and employ expert teams to achieve similar insights over months. Google’s AI accomplished this in significantly less time.

Expert Perspectives

Dr. Sunil Laxman, systems biologist from Bengaluru, acknowledged the achievement while providing context:

“It’s a nice result and was a well-chosen problem to test the capabilities of an LLM. This would have taken a focussed team of dedicated researchers several months to suggest such a use of the drug.”

However, Dr. Laxman noted that the model didn’t suggest anything beyond trained biologist capabilities, emphasizing that most Indian labs lack access to extensive compound libraries for testing.

Mathematical Reasoning Potential

Professor Siddhartha Gadgil from IISc Bengaluru sees broader implications, comparing current AI capabilities to skilled mathematicians with potential for solving complex problems like the Riemann Hypothesis.

“We can’t say when an AI model will solve the Riemann Hypothesis but there’s no reason to suppose it never can,” said Prof. Gadgil.

He cited OpenAI’s experimental reasoning model that performed at gold-medal level in the International Mathematical Olympiad 2025 without specific training.

Research Community Response

The mathematics community remains divided on AI integration, though Prof. Gadgil advocates for embracing these tools, noting their unexplored latent capabilities and potential to accelerate research.