Overcoming confounding plate effects in differential expression analyses of single-cell RNA-seq data
Overcoming confounding plate effects in differential expression analyses of single-cell RNA-seq data
复制标题
DOI：
10.1093/biostatistics/kxw055
复制
发表时间：
2017-07-01
期刊：
影响因子：
2.1
通讯作者：
Marioni, John C.
中科院分区：
文献类型：
--
作者：
Lun, Aaron T. L.;Marioni, John C.
越来越多的研究正在使用单细胞RNA测序(scRNA-seq)来表征单个细胞的基因表达谱。一种应用于scRNA-seq数据的常见分析涉及检测不同生物群中的细胞之间的差异表达(DE)基因。然而，许多实验的设计使得要比较的细胞在单独的平板或芯片中进行处理，这意味着分组与系统的平板效应相混淆。在对scRNA-seq数据的DE分析中，这一令人困惑的方面常常被忽视。在这篇文章中，我们证明了在统计模型中没有考虑板块效应会导致I类错误控制的丢失。提出了一种解决方案，将每个平板中所有单元格的计数相加，并在DE分析中使用所有平板的计数和。这在存在平板效应的情况下恢复了I类差错控制，而不会损害模拟数据中的检测能力。求和对每个板上不同数量和库大小的细胞也是稳健的。在对真实数据的DE分析中也观察到了类似的结果，在这些分析中，使用计数和而不是单细胞计数提高了相关基因的特异性和排名。这表明，在具有平板效应的scRNA-seq数据的DE分析中，求和有助于保持统计的严谨性。
An increasing number of studies are using single-cell RNA-sequencing (scRNA-seq) to characterize the gene expression profiles of individual cells. One common analysis applied to scRNA-seq data involves detecting differentially expressed (DE) genes between cells in different biological groups. However, many experiments are designed such that the cells to be compared are processed in separate plates or chips, meaning that the groupings are confounded with systematic plate effects. This confounding aspect is frequently ignored in DE analyses of scRNA-seq data. In this article, we demonstrate that failing to consider plate effects in the statistical model results in loss of type I error control. A solution is proposed whereby counts are summed from all cells in each plate and the count sums for all plates are used in the DE analysis. This restores type I error control in the presence of plate effects without compromising detection power in simulated data. Summation is also robust to varying numbers and library sizes of cells on each plate. Similar results are observed in DE analyses of real data where the use of count sums instead of single-cell counts improves specificity and the ranking of relevant genes. This suggests that summation can assist in maintaining statistical rigour in DE analyses of scRNA-seq data with plate effects.