FlowSorted.BloodExtended.EPIC: An Enhanced DNA Methylation Library for Deconvoluting Peripheral Blood

DNA methylation microarrays have been extensively used for understanding cell type composition in complex tissue samples. Here we expand on existing libraries for reference-based deconvolution of blood DNA methylation data assayed using the Illumina HumanMethylationEPIC array to include 12 different leukocyte subtypes (neutrophils, eosinophils, basophils, monocytes, B cells naïve and memory, CD4+ and CD8+ naïve and memory cells, natural killers, and T regulatory cells). Application of the IDOL algorithm for identifying optimal libraries for the deconvolution of blood-derived DNA methylation data led to to a library consisting of 1200 CpGs. The accuracy of deconvolution estimates obtained using our IDOL-optimized library was evaluated using artificial mixtures with known cellular composition and in samples were both whole-blood DNA methylation data and FACS information were available. We further showcase potential applications of our expanded library using publicly available cancer, aging, and autoimmune disease data sets. Bisulphite converted DNA from neutrophils (Neu, n=6), eosinophils (Eos, n=4), basophils (Bas, n=6), monocytes (Mono, n=5), B lymphocytes naive (Bnv, n=4), B lymphocytes memory (Bmem, n=6), T helper lymphocytes naive (CD4nv, n=5), T helper lymphocytes memory (CD4mem, n=4), T regulatory cells (Treg, n=3), T cytotoxic lymphocytes naive (CD8nv, n=5), T cytotoxic lymphocytes memory (CD8mem, n=4), and natural killer lymphocytes (NK, n=4), and 12 DNA artificial mixtures (labeled as MIX in the dataset) were hybridised to the Illumina Infinium HumanMethylationEPIC Beadchip v1.0_B4