RT Journal Article
SR Electronic
T1 SCPattern: A statistical approach to identify and classify expression changes in single cell RNA-seq experiments with ordered conditions
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 046110
DO 10.1101/046110
A1 Ning Leng
A1 Li-Fang Chu
A1 Jeea Choi
A1 Christina Kendziorski
A1 James A. Thomson
A1 Ron Stewart
YR 2016
UL http://biorxiv.org/content/early/2016/03/29/046110.abstract
AB Motivation With the development of single cell RNA-seq (scRNA-seq) technology, scRNA-seq experiments with ordered conditions (e.g. time-course) are becoming common. Methods developed for analyzing ordered bulk RNA-seq experiments are not applicable to scRNA-seq, since their distributional assumptions are often violated by additional heterogeneities prevalent in scRNA-seq. Here we present SC-Pattern - an empirical Bayes model to characterize genes with expression changes in ordered scRNA-seq experiments. SCPattern utilizes the non-parametrical Kolmogorov-Smirnov statistic, thus it has the flexibility to identify genes with a wide variety of types of changes. Additionally, the Bayes framework allows SCPattern to classify genes into expression patterns with probability estimates.Results Simulation results show that SCPattern is well powered for identifying genes with expression changes while the false discovery rate is well controlled. SCPattern is also able to accurately classify these dynamic genes into directional expression patterns. Applied to a scRNA-seq time course dataset studying human embryonic cell differentiation, SCPattern detected a group of important genes that are involved in mesendoderm and definitive endoderm cell fate decisions, positional patterning, and cell cycle.Availability and Implementation The SCPattern is implemented as an R package along with a user-friendly graphical interface, which are available at:https://github.com/lengning/SCPatternContact: rstewart{at}morgridge.org