On the discovery of population-specific state transitions from single-cell RNA sequencing data

Single-cell RNA-sequencing (scRNA-seq) has quickly become an empowering technology to profile the transcriptomes of individual cells on a large scale. Many early analyses of differential expression have aimed at identifying differences between cell types (or clusters), and thus are focused on finding markers for cell populations either in a single sample or across multiple samples. More generally, such methods can compare expression levels in multiple sets of cells, thus leading to cross-condition analyses. However, given the emergence of replicated multi-condition scRNA-seq datasets, an area of increasing focus is making sample-level inferences, termed here as differential state analysis. For example, one could investigate the condition-specific responses of specific immune cell subsets across cells measured from patients within each condition, however, it is not clear which statistical framework best handles this situation. In this work, we surveyed the methods available to perform cross-condition differential state analyses, including cell-level mixed models and methods based on aggregated ``pseudobulk'' data. We developed a flexible simulation platform that mimics both single and multi-sample scRNA-seq data and provide robust tools for multi-condition analysis within the R package.

单细胞RNA测序（Single-cell RNA-sequencing, scRNA-seq）已迅速成为一种强大的技术，可在大规模尺度下解析单个细胞的转录组。早期的诸多差异表达分析研究多致力于识别细胞类型（或细胞簇）之间的表达差异，因此其核心目标多为在单一样本或跨多样本的场景下筛选细胞群体的标记基因。更一般地，此类方法可对多组细胞的表达水平进行比较，从而实现跨条件分析。然而，随着重复多条件scRNA-seq数据集的涌现，学界日益关注的研究方向转向样本水平的推断，本文将其称为差异状态分析（differential state analysis）。例如，研究者可针对每个条件下从患者样本中测得的细胞，探究特定免疫细胞亚群的条件特异性响应，但目前尚不明确哪种统计框架最适配此类分析场景。本研究对可用于开展跨条件差异状态分析的现有方法进行了系统综述，涵盖细胞水平混合模型以及基于聚合‘伪批量（pseudobulk）’数据的分析方法。我们开发了一款灵活的模拟平台，可模拟单样本和多样本的scRNA-seq数据，并在R包中集成了适用于多条件分析的稳健分析工具。