Cancer Grand Challenges launches AURORA to test AI co-scientists in cancer labs

Cancer Grand Challenges said on Thursday, Nov. 6, 2026, that Project AURORA will put AI co-scientists to work alongside three pioneer teams — NexTGen, PROMINENT, and MATCHMAKERS — to generate hypotheses and prioritize experiments in cancer biology, with researchers retaining final scientific decisions.

Cancer labs drown in data, and sorting which experiment to run still depends on scarce expert hours. AURORA's bet is software agents that propose and rank experiments, then get graded against what the lab bench actually finds, without handing the steering wheel to the model. $10 million is not a moonshot cure budget. It is a controlled test of whether co-scientists speed real challenges or just generate prettier wrong answers.

Cancer Grand Challenges said on Nov. 6, 2026 that it is launching Project AURORA, led by Marinka Zitnik at Harvard University. Cancer Research UK is funding it through Cancer Grand Challenges, at $10 million. The job is to put AI co-scientists beside teams that already hold a cancer challenge. A co-scientist, here, is software that can read the evidence, suggest a hypothesis, and propose an experiment. A hypothesis is a specific claim a lab can try to knock down. The researcher still decides what counts. Those lines are Cancer Grand Challenges'.

Three pioneer teams will receive a combined $4 million, on top of the funding they already have, to move faster on their own challenges. The three are NexTGen, PROMINENT, and MATCHMAKERS. The announcement does not divide that $4 million into three separate checks. Those lines are Cancer Grand Challenges'.

NexTGen is developing engineered T-cell therapies for children with solid tumors. A T cell is an immune cell. CAR-T is a treatment that reprograms a patient's own T cells so they can recognize a cancer. A solid tumor is a cancer that forms a mass. Designing one of these treatments is a stack of choices: which feature of the tumor to aim at, how to keep the T cells working, and how to get past the tissue around the tumor that can shut an immune attack down. Nicky McGranahan at University College London leads the AURORA work on NexTGen, using the software to help rank CAR-T designs. Martin Pule, a co-team lead of NexTGen, is in the launch photograph. Those lines are Cancer Grand Challenges'.

PROMINENT is asking why a cell can carry mutations linked to cancer and still behave like normal tissue, and what pushes some of those cells toward a tumor. Nuria Lopez-Bigas at the Institute for Research in Biomedicine in Barcelona leads that team. They plan to test the hypotheses they choose with automated genetic screens and with pictures of cells. A screen, here, is a sweep through many genetic changes to see which ones matter. MATCHMAKERS is trying to predict how T cells recognize cancer. The menu is enormous: which pieces of a tumor to test, and which receptors on a T cell might bind them. Welles Robinson and Sergio Quezada at University College London will work with Zitnik's group on two co-scientists aimed at that problem. Those lines are Cancer Grand Challenges'.

AURORA builds on ToolUniverse, software from Zitnik's group that connects an AI system to scientific tools and data. A co-scientist is supposed to analyze the evidence, write a hypothesis a lab can test, and propose experiments that fit what that program can actually run: the tools, the samples, and the question the team is set up to ask. Zitnik said the goal is not another chatbot that can search the internet. It is to equip these systems with tools a human researcher would use, and to have them work with the scientists. That quotation is hers.

Researchers set the questions, review the hypotheses, and test the ideas they pick in the lab. What the experiment finds goes back into the system, which is supposed to make the next proposal better. The model does not get the last word. Zitnik said cancer research already produces vast and growing amounts of data, and that turning it into scientific insight is still the hard part. She said AURORA is about bringing evidence together and testing new ideas "while keeping researchers firmly at the centre of the discovery process." That quotation is hers.

The launch event was held on Nov. 5, 2026, at Harvard University. The picture, left to right as Cancer Grand Challenges labels it, is Marinka Zitnik, project lead of AURORA; Martin Pule, co-team lead of NexTGen; Nuria Lopez-Bigas, co-team lead of PROMINENT; and Welles Robinson of MATCHMAKERS. They stand beside a Cancer Grand Challenges banner. The credit is Cancer Grand Challenges.

In plain terms, Cancer Research UK is spending $10 million to see whether software can propose cancer experiments that three existing teams will actually run. Those teams share $4 million more. The announcement does not describe a new drug, a finished clinical trial, or a scientist who has been replaced.