How well do RNA-Seq differential gene expression tools perform in a eukaryote with a complex transcriptome?

RNA-seq experiments are usually carried out in three or fewer replicates. In order to work well with so few samples, Differential Gene Expression (DGE) tools typically assume the form of the underlying distribution of gene expression. A recent highly replicated study revealed that RNA-seq gene expression measurements in yeast are best represented as being drawn from an underlying negative binomial distribution. In this paper, the statistical properties of gene expression in the higher eukaryote Arabidopsis thaliana are shown to be essentially identical to those from yeast despite the large increase in the size and complexity of the transcriptome: Gene expression measurements from this model plant species are consistent with being drawn from an underlying negative binomial or log-normal distribution and the false positive rate performance of nine widely used DGE tools is not strongly affected by the additional size and complexity of the A. thaliana transcriptome. For RNA-seq data, we therefore recommend the use of DGE tools that are based on the negative binomial distribution.

RNA - seq实验通常在三个或更少的重复样本中进行。为了在如此少的样本情况下良好运行，差异基因表达（DGE）工具通常假定基因表达的潜在分布形式。最近一项高度重复的研究表明，酵母中的RNA - seq基因表达测量值最好被表示为来自潜在的负二项分布。在本文中，尽管转录组的大小和复杂性大幅增加，但高等真核生物拟南芥的基因表达统计特性与酵母的基本相同：来自这种模式植物物种的基因表达测量值与来自潜在的负二项分布或对数正态分布相符，并且九种广泛使用的DGE工具的假阳性率性能不受拟南芥转录组额外的大小和复杂性的强烈影响。因此，对于RNA - seq数据，我们建议使用基于负二项分布的DGE工具。