A team from the Universidad Politécnica de Madrid (UPM) and the Universidad de Hamburgo (Germany) has managed to unravel how Alzheimer’s affects several types of brain cells. To do so, they analyzed 1.3 million cells belonging to 18 different cell classes, using advanced network medicine techniques and artificial intelligence (AI).
"Although its visible signals in the brain are well known, Alzheimer’s still holds many mysteries because it remains unclear what happens within each type of brain cell: which genes become dysregulated, who directs them, and why some neurons are more vulnerable than others," the researcher Andrea Álvarez explained.
The specialists emphasize the relevance of the approach used to study brain cells, comparing samples from healthy individuals with samples from people with Alzheimer’s taken from the public atlas "Seattle Alzheimer's Disease Brain Cell Atlas" (SEA-AD).
In this work, network medicine has been integrated, which models genes as a network to identify which ones act in a coordinated manner and which "switches" regulate their activity, with the scGPT AI model, trained with more than 33 million human cells in a specific version for brain tissue.
The findings indicate that Alzheimer’s-linked alterations group into specific sets of genes associated with inflammation, communication between neurons, and the matrix surrounding the cells. Both tools point to the same key genes: the AI highlights which ones are more influential, while network medicine details how they relate to one another.
"The most outstanding finding occurred in SST neurons, a type of neuron particularly vulnerable to Alzheimer’s. In them, a new possible regulatory mechanism was identified that links inflammation and cellular damage, and that is precisely altered in the most affected neurons," Álvarez explained, a UPM doctoral student who developed this work during a stay at the Institute for Computational Systems Biomedicine in Hamburg.
Practical implications and future lines of research
Although these results will need to be validated experimentally in the laboratory, the researchers point out that they provide key information: knowing which mechanisms are modified in each type of cell can clarify why some neuronal populations show greater vulnerability to Alzheimer’s and guide the design of new lines of research into potential therapeutic targets.
The work does not yet propose a treatment or immediate clinical applications, but the authors emphasize the knowledge generated, useful for improving early detection and exploring new therapeutic strategies against a condition that affects more than 800,000 people in Spain and for which around 40,000 new cases are diagnosed each year, according to the Sociedad Española de Neurología (SEN).
The study also provides a computational roadmap to prioritize genes, interpret their interactions, and put forward new hypotheses about the molecular mechanisms involved in the disease, showing that the combination of artificial intelligence and network medicine makes it possible not only to identify which genes are relevant, but also to understand the reasons for their importance.
With this, the role of AI and computational biology is consolidated as essential tools for deepening the study of neurodegenerative diseases and moving toward a more detailed vision of their biological complexity.