Abstract:
|
Cells are the basic biological units of multicellular organisms. Recent breakthroughs in single-cell RNA sequencing (scRNA-seq) have made possible the profiling of gene expression at the single-cell level, paving the way for exploring gene expression heterogeneity among cells. In this talk, I will explore the noise properties of unique molecular identifier (UMI)-based scRNA-seq data, and show, by comparisons to Fluorescent in situ RNA hybridization (RNA FISH), that a simple Poisson-based model is sufficient for recovering true underlying gene expression distributions using a deconvolution method called DESCEND. I will also describe SAVER, a gene expression recovery framework for scRNA-seq data. SAVER reduces the noise in the observed scRNA-seq data and "recovers" the expression concentrations for genes with zero read counts due to low efficiency sampling. I will illustrate how SAVER affects downstream analyses such as cell type clustering, differential expression analysis, and the assessment of gene-gene relationships.
|