Humanmethylation450 450k microarray platform

We implemented in silico estimation of the relative proportions of three cell types (ES-derived NPCs from culture ^{74 (link)}, and adult cortex neuronal and non-neuronal cells from tissue ⁷⁵ ) using epigenome-wide DNAm data using a recently published algorithm ^{76 (link)}. All data was obtained using the Illumina HumanMethylation450 (“450k”) microarray platform. After normalizing the publicly available data together using the preprocessQuantile function in the minfi Bioconductor package ⁷⁷ , we picked the cell type-discriminating probes as outlined by Jaffe and Irizarry ^{78 (link)} resulting in 227 unique probes that distinguished the three cell types. We then normalized the DNAm data from our 36 discovery samples, and estimated the composition of our samples from the methylation profiles of the homogenate cell types at the 227 probes using non-linear mixed modeling ^{76 (link)}. Composition estimates were regressed against the normalized and log₂ transformed expression levels within each DER across the 36 samples, and we obtained a moderated T-statistic and corresponding p-value ⁷⁹ for each cell type and DER. The Bonferroni–adjusted p-value was set at 0.05/50,560, or p < 9.89×10⁻⁷.