Cancer researchers start testing AURORA AI co-scientists built at Harvard

Three teams from Cancer Grand Challenges will test AURORA, an agentic AI system designed to analyze research evidence, generate hypotheses and suggest experiments. Cancer Research UK detailed the test on its website as the next stage of the £8 million project.

A team led by Marinka Zitnik at Harvard Medical School developed the system. AURORA can examine scientific literature, research databases and experimental findings, while providing reasoning traces for its suggestions. Human researchers retain control over scientific decisions and will test its proposals in the laboratory. Zitnik's team received a grant of nearly $7 million from Cancer Grand Challenges to build the co-scientists, and the project includes an additional $4 million for three teams in Spain, the UK and the US, according to Harvard Medical School.

The NexTGen team will study ways to modify immune cells so they can find and attack solid tumors in children and young people. PROMINENT will examine why some cells with cancer linked genetic changes remain healthy while others become cancerous. MATCHMAKERS will investigate how T cells recognize cancer and which cancer markers could provide suitable treatment targets.

The teams will compare AURORA's recommendations with experimental results and provide feedback to improve the system. The evaluation will measure whether its proposed findings hold up under laboratory testing.