Preprint authors combined 13 brain studies and identified four molecular profiles in samples from people with Alzheimer’s disease

Preprint authors combined 13 brain studies and identified four molecular profiles in samples from people with Alzheimer’s disease

On September 4, the authors posted a preprint describing the panAD atlas. They harmonized data from more than three million cell nuclei obtained from postmortem cortical samples from 791 people. Based on gene activity across seven major cell types, the authors identified four exploratory groups among samples from people diagnosed with Alzheimer’s disease.

Single-cell studies of Alzheimer’s disease are difficult to compare because they examine different cortical regions and participant groups and use different tissue preparation methods. Agreement between their results could therefore reflect either shared disease biology or characteristics of a particular participant group. The authors processed 13 datasets using a common pipeline to determine which combinations of active genes recurred across studies.

The authors used 391 samples from people diagnosed with Alzheimer’s disease to define the four profiles. For each sample, the MONET model integrated gene activity data from seven cell types and adjusted the data for age, sex, and the original dataset. Twenty model runs produced similar sample assignments. A statistical test of group separation provided stronger support for two or three clusters. The authors selected four clusters to examine finer differences. In the paper, these four profiles serve as exploratory labels for comparing samples.

In the subset of 217 samples with complete data, the profiles had broadly similar densities of amyloid plaques, which are protein deposits assessed in Alzheimer’s disease. The profiles differed in the extent of pathological tau protein and in scores on a brief test of memory and thinking. The AD1 profile had a lower stage of tau pathology than the other profiles and the lowest test scores. AD2 and AD4 had higher scores despite a high pathological burden.

Postmortem samples from people with the same diagnosis showed different combinations of cellular programs, pathological changes, and cognitive outcomes. The atlas gives future studies a way to relate disease mechanisms and potential interventions to these combinations.