Researchers led by the University of Cambridge have used artificial intelligence (AI) tools to find new ways to treat cancer using drugs that are already safe and affordable. They used GPT-4 to search through scientific studies and find patterns that humans might miss. Their goal was to identify drug combinations that could fight breast cancer without harming healthy cells. They focused on drugs that are cheap and already approved for other conditions, like high cholesterol or alcohol dependence.
The researchers gave GPT-4 specific instructions to suggest drug pairs that could target breast cancer cells in a lab setting. They told the AI to avoid typical cancer drugs and focus on safe, affordable options. After GPT-4 made its suggestions, human scientists tested these drug combinations in the lab to see how well they worked against breast cancer cells. In the first round of tests, three out of 12 drug pairs suggested by GPT-4 performed better than existing breast cancer treatments. The AI then learned from these results and suggested four more combinations, three of which also showed good results.
A new way to collaborate with AI
The study, published in the Journal of the Royal Society Interface, showed a new way of working where lab results guide the AI, and the AI’s ideas guide further experiments. This back-and-forth process is called a closed-loop system. The researchers explained that GPT-4 isn’t replacing scientists but acting as a helpful partner. Sometimes, the AI makes mistakes, but in this case those mistakes were useful because they led to new ideas worth testing.
Among the drug pairs GPT-4 suggested, two drugs stood out: simvastatin, which is used to lower cholesterol, and disulfiram, which helps treat alcohol dependence. These drugs showed promise against breast cancer cells in the lab. However, they would need more testing in clinical trials before they can be used to treat cancer in people. The researchers believe this approach could speed up drug discovery by exploring many possibilities quickly. They see AI as a tool that can suggest new ideas and help scientists find treatments faster, especially in complex areas like cancer research.