Cell-type expression specificity

We applied the EWCE R-package (https://github.com/NathanSkene/EWCE) reported by Zeisel’s study¹⁶ to investigate cell-type expression specificity of AD-related genes. The EWCE method was demonstrated to be a feasible approach to study the expression specificity of a gene list across several different cell types with single-cell transcriptomes^14,15. The EWCE method employs various single-cell transcriptome datasets^17–20 from mice brain regions, including neocortex, hippocampus, hypothalamus, striatum, and midbrain. These data were generated by an identical method in Karolinska Institutet and were observed with no important batch effects. A total of 9970 cells were merged into a matrix¹⁴ that annotates 24 brain cell types (e.g., pyramidal neurons, interneurons, oligodendrocytes, astrocytes, microglia, vascular endothelial cells, mural cells, and ependymal cells). Cell types were identified via a backspin algorithm described in corresponding studies associated with the dataset. The EWCE method calculates the average expression level of gene in each cell type and then calculates the specificity of gene in each cell type. The specificity is calculated by the mean expression in one cell type divided by the mean expression in all cell types. For a list of target genes, EWCE calculates the cell-type specificity of target genes and then estimates the P-value of specificity of target genes compared with the specificity of background genes via a bootstrap method. This bootstrap method randomly samples 10,000 gene lists with the same number of target genes from all the genes as background genes, and then estimates the distribution of specificity of background genes. P-values of specificity from multiple tests were adjusted by the false discovery rate (FDR) method.