Fig. 1

Construction and validation of the 12 blood cell-type hypoDNAm reference matrix. a Left panel: Example of a CpG’s DNAm profile in the DNAm reference matrix, marking T-regulatory (Treg) cells. The y-axis labels the cell types, x-axis labels the DNAm value, and the number of samples of each cell type (i.e., in each boxplot) is shown on the y-axis. Right panel: The DNAm reference matrix for 12 blood cell subtypes encompassing 600 CpGs (i.e., 50 markers per cell type). b Scatterplots of true fractions vs estimated fractions for 10 blood cell subtypes using the EPIC DNAm data from 10 artificial mixtures where the underlying mixing proportions were known. For each estimated cell type, we display the R-value (Pearson correlation coefficient) and root mean square error (RMSE). c Heatmap displays the estimated fractions of cell-sorted samples for each of the 12 immune-cell subtypes in our Illumina 450k/850k DNAm reference matrices, as well as the total CD4 + T-cell, total CD8 + T-cell, and total B-cell fractions. The immune-cell type of the sorted sample is indicated by the color bar on top of the heatmap. The study from which the sorted sample derives from is indicated by the color bar below the heatmap. The technology used to generate the DNAm data of the sorted sample is also indicated. For the 450k and WGBS samples, we used the 450k and 850k DNAm reference matrices, respectively, to obtain the fractions. The estimated fractions in the heatmap are median values taken over biological replicates of the cell-sorted samples, with the number of corresponding biological replicate samples indicated at the bottom