What if Alzheimer’s disease is not one illness following one path, but several different biological problems hiding under the same diagnosis? This study found three distinct patterns in affected brains. Alzheimer’s disease is the most common type of dementia in older people.

It has become a significant public-health challenge, affecting both lives and the cost of care. That matters because Alzheimer’s disease and dementia impose a much heavier financial burden: average medical-insurance service costs are more than three times those for other older adults.

But Alzheimer’s disease is a heterogeneous disease: different people can have different biological changes beneath the same diagnosis. This inconsistency among molecular subtypes has led to diagnostic ambiguity and failure in drug development. The study therefore distinguished patients at the level of their gene activity.

To look beneath the diagnosis, the study examined gene expression data and clinical features from brain tissue, using one thousand two hundred forty samples, including Alzheimer’s and non-dementia samples. The information came from three independent studies, allowing the researchers to compare patterns across separate collections of brain samples.

So they used consensus clustering to classify the Alzheimer’s samples into different subgroups, asking whether the dataset could be separated into more than one group. The goal was to use consensus clustering to classify the Alzheimer’s samples into different subgroups within the Alzheimer’s disease dataset.

The gene activity patterns consistently separate the Alzheimer’s samples into three groups, with each group showing a stability score above zero point eight. That gives the later comparisons a dependable basis: these are recurring biological patterns, not arbitrary divisions.

The three-subgroup classification was more robust than the other classifications considered, and each subgroup had a high consensus score. The three groups were therefore selected for later analysis. The gene-expression patterns were highly similar within each subgroup and significantly different between subgroups.

The three subgroups differed in their clinical and biological features. Compared with people without Alzheimer’s disease, all three had higher measures linked to disease-related brain changes, and their brain tissue had lower pH.

This means the groups shared signs of Alzheimer’s disease, but they were not identical in how those signs appeared. The three subgroups all carried more severe signs of Alzheimer’s disease than the comparison group, but they were not identical: Subgroups I and III had the strongest disease-related changes, while Subgroup I also had more nerve-cell loss.

The study then checked whether combinations of eight genes could help identify the subgroups in a separate dataset. The combination was the best diagnostic scheme tested for each subgroup. Its diagnostic value was strongest for subgroup one, followed by subgroup two and then subgroup three, suggesting that these gene patterns could help distinguish the groups.

But these results show associations, not causes. Data-driven biomarkers do not directly account for all the many genes and other factors that influence Alzheimer’s disease. The study also notes that testing the subtypes in an animal model is particularly difficult to carry out in practice.

It also warns that some important information may remain hidden when the analysis examines only the disease-related portion of the findings. The study combined multiple unsupervised subtype maps into biological subtypes with clinical predictive value in the analysis of Alzheimer’s disease.

The study emphasizes that exploration after intervention is required, a requirement it identifies for moving this line of investigation forward. The study suggests that separating Alzheimer’s into biological subgroups could make diagnosis more precise and point toward more personalized treatment, although these patterns still need testing to show cause and effect.