| Abstract: |
| Gene expression levels vary greatly from cell to cell, leading to significant consequences in many biological process from bacterial decision-making to mammalian development. The underlying processes responsible for generating expression variability are poorly understood. Single-cell RNA-seq provides an unprecedented opportunity to decipher this phenomenon, and statistical methods need to be developed to interpret stochasticity in gene expression. We propose a scalable computational pipeline (BMA) to infer the kinetics of stochastic gene expression from single-cell RNA-seq data. Given an underlying model of gene expression, BMA uses a binomial moments approximation method to identify predictive models of transcriptional dynamics. We generate gene transcription models with varying complexity and use BMA to select the most predictive model, then apply BMA to single-cell RNA-seq data of hematopoietic progenitor cells. These results illustrate that BMA provides a flexible and efficient way to investigate the kinetics of transcription. |
|