Reliability and correlation of mixture cell correction in methylomic and transcriptomic blood data.

Objectives: The number of DNA methylome and RNA transcriptome studies is growing, but investigators have to consider the cell type composition of tissues used. In blood samples, the data reflect the picture of a mixture of different cells. Specialized algorithms can address the cell-type heterogeneity issue. We tested if these corrections are correlated between two heterogeneous datasets.
Results: We used methylome and transcriptome datasets derived from a cohort of ten individuals whose blood was sampled at two different timepoints. We examined how the cell composition derived from these omics correlated with each other using "CIBERSORT" for the transcriptome and "estimateCellCounts function" in R for the methylome. The correlation coefficients between the two omic datasets ranged from 0.45 to 0.81 but correlations were minimal between two different timepoints. Our results suggest that a posteriori correction of a mixture of cells present in blood samples is reliable. Using an omic dataset to correct a second dataset for relative fractions of cells appears to be applicable, but only when the samples are simultaneously collected. This could be beneficial when there are difficulties to control the cell types in the second dataset, even when the sample size is limited.