Fig. 2

General analytical workflow for large-scale single-cell/nucleus RNA-seq data from human control and AD samples. A Starting from the genes × cells/nuclei counts table, most analysis workflows identify high variance genes, then perform dimensionality reduction, and ultimately call clusters in reduced-dimension space. This clustering may be iterative, where larger clusters are then re-analyzed, starting from the first step, to identify subgroups (enlarged inset on the right). It is important to note that often a separate reduced-dimension embedding is used for visualization, as opposed to the embedding used for clustering. B–D After clusters have been identified, the analysis workflow looks to identify differences in gene expression across conditions in each major cell class (B) or subcluster (C) or in the relative proportions of each cell class or cluster across conditions (D). These differences form the basis of understanding cell type-specific changes associated with the disease