Abstract:
|
Association analysis of DNA methylation (DNAm) data is challenged by cell type heterogeneity, as the data is typically a mixture of cell types. Cell type heterogeneity can bias results as DNAm is a mechanism in tissue and cell differentiation. To correct for this, analyses include observed or estimated cell type counts as covariates. This does not estimate exposure effects on cell type specific DNA methylations. Direct measurements of cell type specific methylation would involve costly lab work. We developed a statistical method to estimate cell specific associations using whole blood methylation data when cell composition is available but cell-specific methylations are not. We assume cell type specific regression models of the exposure effects on cell type specific methylations. We then relate these models to the observed data and use GEE2 estimation for cell-specific effects. We analyzed data from the Normative Aging Study for cell specific smoking associations on 49 probes established to be associated with smoking. Five probes had a significant cell type specific association. To the best of our knowledge, this is the first method to estimate these effects from whole blood.
|