RT Journal Article
SR Electronic
T1 Individual Level Differential Expression Analysis for Single Cell RNA-seq data
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 2021.05.10.443350
DO 10.1101/2021.05.10.443350
A1 Mengqi Zhang
A1 Si Liu
A1 Zhen Miao
A1 Fang Han
A1 Raphael Gottardo
A1 Wei Sun
YR 2021
UL http://biorxiv.org/content/early/2021/05/10/2021.05.10.443350.abstract
AB Bulk RNA-seq data quantify the expression of a gene in an individual by one number (e.g., fragment count). In contrast, single cell RNA-seq (scRNA-seq) data provide much richer information: the distribution of gene expression across many cells. To assess differential expression across individuals using scRNA-seq data, a straightforward solution is to create “pseudo” bulk RNA-seq data by adding up the fragment counts of a gene across cells for each individual, and then apply methods designed for differential expression using bulk RNA-seq data. This pseudo-bulk solution reduces the distribution of gene expression across cells to a single number and thus loses a good amount of information. We propose to assess differential expression using the gene expression distribution measured by cell level data. We find denoising cell level data can substantially improve the power of this approach. We apply our method, named IDEAS (Individual level Differential Expression Analysis for scRNA-seq), to study the gene expression difference between autism subjects and controls. We find neurogranin-expressing neurons harbor a high proportion of differentially expressed genes, and ERBB signals in microglia are associated with autism.Competing Interest StatementThe authors have declared no competing interest.