A team of researchers at the University of Cambridge has discovered that combinations of widely available non-cancer drugs, such as those used to treat high cholesterol and alcohol dependence, could be effective in killing cancer cells.
The study used a large language model (LLM) called GPT-4 to identify potential new drug combinations. The LLM was prompted to suggest pairs of drugs that would target breast cancer cells without harming healthy ones, and also prioritize affordable options approved by regulators.
Three out of 12 suggested combinations worked better than existing breast cancer treatments in lab tests, with the LLM identifying four more promising combinations that showed similar results.
The researchers say that supervised AI tools like GPT-4 can aid human scientists in areas such as drug discovery, rather than replacing them. The collaboration between humans and AI led to unconventional but potentially valuable ideas for new cancer treatments.
Some of the suggested combinations, including simvastatin (used to lower cholesterol) and disulfiram (used in alcohol dependence), showed promise against breast cancer cells. However, further clinical trials would be needed before these treatments could be used in cancer care.
The study highlights the potential for AI-powered tools to accelerate scientific discovery and generate new ideas in areas like cancer research.
Source: https://medicalxpress.com/news/2025-06-ai-scientist-combinations-widely-cancer.html