A team of applied mathematicians has built a whole-brain computer model that simulates, region by region, how the buildup of amyloid-beta protein and reductions in blood flow interact over time, according to a paper published in Computer Methods in Applied Mechanics and Engineering. The model, described first as a preprint on arXiv and then in the peer-reviewed journal, combines equations for how amyloid-beta is produced, spreads and is cleared with a separate set of equations modeling blood flow through arteries, capillaries and veins, then couples the two systems together in simulated brain geometries. The researchers, based at Politecnico di Milano, the University of Vienna, the University of Oxford and Friedrich-Alexander-Universitat Erlangen-Nurnberg, say the work is a mathematical simulation, not a study of human tissue or living patients.
The central finding is what the authors call multistability: in the simulations, a small, localized patch of amyloid-beta could be cleared away, letting the modeled brain tissue return to a normal state, while a larger starting patch instead triggered a self-sustaining spread across the simulated brain. When the researchers additionally simulated a localized loss of blood flow, such as might follow a stroke-like event, smaller amounts of amyloid that the model could otherwise clear became enough to trigger wider spread. The result offers a mechanistic account, built entirely from equations and simulated anatomy, for why regions with vascular damage and regions with heavy amyloid deposits tend to show up together on scans of people with Alzheimer's disease.
The idea that amyloid and blood vessel damage reinforce each other is not new. Earlier laboratory work, including studies of genetically engineered mice and of amyloid's effects on human brain capillary cells, had already suggested that amyloid-beta can constrict small vessels and that reduced blood flow can in turn promote amyloid production. What the new paper adds is a way to run that loop forward in time across an entire simulated brain and see which starting conditions lead to clearance versus runaway spread. The authors themselves frame the model's value as prospective: they suggest tools like this one might eventually help identify risk conditions or guide prevention research, language that describes a hope for future use, not a demonstrated clinical benefit.
Outside experts were not quoted in the materials reviewed, and the paper itself is presented by its authors as a numerical and mathematical proof of concept rather than a validated diagnostic or prognostic tool.