et al.). We evaluated batch effects based on the clustering results and visual inspection of the UMAP plots, where none of the samples clustered separately in any of the clusters, which is, in this case, sufficient to obtain cell types. We clustered the data using Louvain clustering with resolution 1 and annotated cell types using the previously reported marker genes as microglia 1 (CTSD, CD9, HEXB, CST3), microglia 2–3 (LPL, CST), granulocytes (CAMP, S100a9), T/NK cells (S100a4, NKG7, Trbc2), B cells (RAG1, CD79b, CD74), monocytes (S100a4, CD74), perivascular macrophages (CD74, CD163, MRC1) (see Supplementary Fig. 6). We subsequently sub-clustered the microglia population into three clusters, assigning the labels microglia 1, 2, and 3, respectively. Similar to Keren-Shaul et al., we assigned the region-sorted samples of AD and WT mouse model (n = 2 per region) with a k-nearest neighbor classifier (k = 30). We then evaluated the number of unassigned cells, performed another round of Louvain clustering, and assigned the remaining cells based on the majority vote for the clustering result, i.e., when unassigned cells clustered predominantly with microglia 1, they were all assigned to microglia 1. The obtained proportions of microglia subpopulations are in accordance with the previously reported
