Fig. 5

Impact of different cellular composition estimates on the analysis of H3K27ac ChIP-seq data. a Hierarchical clustering of the samples based on sample-sample correlation indicates that samples cluster according to their cellular composition, estimated by MSPs. b Association of demographic variables or cell estimates with the first five principal components of the data. The indicated numbers show the p-values of variables’ beta coefficients (y-axis) with the first five PCs (x-axis). c,d Comparison of the Marzi et al. data analysis output based on the different models, with cell type-specific H3K27ac ChIP-seq data from Girdhar et al. [25]. c H3K27ac fold of change in neurons vs glia in regions hypo- or hyperacetylated in AD based on each model. d Density plots of the correlation between the estimated H3K27ac effect sizes based on each model with the H3K27ac effect size in neurons compared to glial cells. Contours of the differentially acetylated regions based on each model are indicated in orange