A team of AI researchers, physicians, and scientists centered at MIT has released FINGERS-7B, described as the first AI foundation model built specifically to make Alzheimer's disease preventable. The model combines lifestyle, clinical, genomic, and proteomic data drawn from tens of thousands of at-risk individuals, with the goal of identifying multi-omic biomarkers that signal preclinical Alzheimer's - the stage before memory and thinking problems become noticeable.
Tested on datasets from the WW-FINGERS network, the model reportedly delivers four times more accurate preclinical diagnosis and 130 percent better responder stratification than previous approaches. That second measure matters as much as the first: responder stratification is about predicting who is likely to benefit from a given prevention strategy, which is the difference between offering everyone the same generic advice and matching individuals to interventions that stand a real chance of working for them.
The practical significance lies in timing. Alzheimer's pathology builds silently for years, and by the time symptoms surface, much of the damage is done. Tools that can identify high-risk individuals during that silent window give researchers a way to test prevention trials in the people most likely to show a measurable effect, and give clinicians a longer runway to act.
FINGERS-7B is open source and has been deployed in the AD Workbench, the secure cloud environment run by the Alzheimer's Disease Data Initiative that Alzheimer's researchers use worldwide - meaning other groups can examine and build on it rather than take the results on faith. The team will present the work at ICLR, one of the largest AI conferences, on April 27, 2026 in Rio de Janeiro.