Abstract:
|
Bulk RNA sequencing provides reliable data for gene expression at the tissue level. A recent technique, single-cell RNA-seq (scRNA-seq), deepens those analyses by evaluating gene expression at the cell type level. However, constrained by current technology and cost, scRNA-seq data are known to be noisy and typically are collected from a small number of subjects, which results in data lack sufficient variability, and this limits their usage. To address these issues while maintaining the unique advantages of each data type, we develop a fully Bayesian deconvolution method (bMIND) to integrate bulk and scRNA-seq data. With scRNA-seq data as a reference, we propose to deconvolve bulk expression data to produce estimates for subject-level cell-type-specific (CTS) expression. The CTS expression enables downstream analyses such as CTS differential expression (DE) analysis. Through simulations, we demonstrate that bMIND improves the accuracy of subject-level CTS expression estimates and the power of discovering DE genes as compared to existing methods. We apply bMIND to data relevant to psychiatric disorders and identify new CTS-DE genes as compared to those derived from scRNA-seq data.
|