A group of major medical institutions specializing in cancer care have formed a partnership to better take advantage of AI’s potential to advance the space. With $40 million of cash and resources from big tech backers, the Cancer AI Alliance (CAIA) could be a huge step forward in precision medicine.
The members of the alliance are Fred Hutchinson, which will coordinate the new effort, Johns Hopkins, Dana Farber, and Sloan Kettering — to be precise, the cancer research arms of these organizations.
As Fred Hutch President and Director Tom Lynch said on stage at the Intelligent Applications Summit in Seattle, where the institute is based, “we believe this has the potential to be transformative. This represents an unprecedented ability… to agree that working together will enable progress.”
He gave the example of a patient with a rare pediatric cancer going at one center, but the scientific knowledge to better treat it is siloed at another center, wrapped in proprietary methods and handling protocols. Perhaps in ten years that knowledge will filter out through the scientific literature, but as he pointed out, the kid with a non-responsive leukemia doesn’t have that long.
AI isn’t some miracle worker, of course, and the tug on the heartstrings isn’t meant to imply that this problem would quickly and easily be solved by some hypothetical treatment-finding model. But if a treatment or study that could help move things forward is not visible between these organizations, it slows down the whole field.
The problem is that sharing data between medical organizations is not simple, due to regulations, safety considerations, and mismatches between formats and databases. Even if the study to help that kid with leukemia at Sloan Kettering is present at Johns Hopkins, there’s no guarantee it will be present in a way that can be shared in a legal and technically feasible way.
The new organization aims to solve this by means of federated learning, a type of secure data collaboration where the raw data stays private, but can be used for the purposes of training AI and other computational systems.
If the research organizations can contribute to a shared goal, like training a drug discovery or diagnostic model for a cancer they all know exists, while complying with HIPAA and other data controls, they will happily do so. Creating a collaborative system under this model is the goal of CAIA, but it’s still a ways out, according to Jeff Leek, VP and Chief Data Officer of Fred Hutch.
It’s certainly possible, he explained, but it’s a difficult problem on the tech side that can only be approached once you have the principal participants in place. Lining up these cancer research centers, and binding them with the money and expertise from Microsoft, AWS, Nvidia, and Deloitte was the necessary first step, and not a trivial one. Now the actual shared infrastructure, standards, and specific goals (such as pursuing a model for a specific cancer or treatment) can begin to take shape.
The $40 million is a mix of operating cash, services, and intangibles from the four companies mentioned, and will be deployed on an unspecified timeline except that CAIA expects to be functional by the end of this year. The initiative should be “producing its first insights” by the end of 2025.
