|
Abstract:
|
The bulk high-throughput omics data contain signals from a mixture of different cell types. Recent developments in deconvolution methods enable one to make cell type specific inferences from the bulk data. Our real data exploration suggests that the differential expression or methylation status are often highly correlated among cell types. Based on this observation, we develop a novel statistical method to account for the cell type hierarchy in cell type specific differential analyses. Extensive simulation and real data analyses demonstrate that incorporating the cell type hierarchy improves the accuracy of detecting the cell type-specific differential signals, especially in cell types with low abundance. Compared with existing methods serving similar purposes, the proposed method achieves significantly better performances.
|
