Zhang Han, Zhang Wei, Zheng Xiaoying, Li Yuanyuan
School of Mathematics and Physics, Wuhan Institute of Technology, Wuhan, 430073, China.
Sci Rep. 2025 May 15;15(1):16963. doi: 10.1038/s41598-025-01969-1.
The remarkable advancement of single-cell RNA sequencing (scRNA-seq) technology has empowered researchers to probe gene expression at the single-cell level with unprecedented precision. To gain a profound understanding of the heterogeneity inherent in cell fate determination, a central challenge lies in the comprehensive analysis of the dynamic regulatory alterations that underlie transcriptional differences and the accurate inference of the differentiation trajectory. Here, we propose the method scRDEN, a robust framework that infers important cell sub-populations and differential expression networks of multiple genes along the differentiation directions of each branch by converting the unstable gene expression values in cells into relatively stable gene-gene interactions (global features) and extracting the order of differential expression (network features), and further integrating the expression features of different dimension reduction methods. When applied to five published scRNA-seq datasets from human and mouse cell differentiation, scRDEN not only successfully captures the stable cell subpopulations with potential marker genes, measures the transcriptional differences of gene pairs to identify the rank differential expression network along the differentiation direction of each branch. In addition, in multiple gene rank differential expression networks, the rank expression directly related to transcription factors/marker genes shows a significant strengthening and weakening trend along with their expression changes, and the distribution of diversity and cluster coefficient show a non-monotonic change trend, including the cases of increasing first and then decreasing or decreasing first and then increasing. This may correspond to the mechanism of cells gradually differentiating into stable functions. It is particularly noteworthy that scRDEN method yielded exceptional results when applied to the large-scale, multi-branched, double-batch mouse dentate gyrus data. This outstanding performance provides novel and valuable insights into large-scale, multi-batch trajectory inference and the study of transcriptional mechanism regulation during the processes of differentiation and development.
单细胞RNA测序（scRNA-seq）技术的显著进步使研究人员能够以前所未有的精度在单细胞水平上探究基因表达。为了深入理解细胞命运决定中固有的异质性，一个核心挑战在于对转录差异背后的动态调控变化进行全面分析，并准确推断分化轨迹。在此，我们提出了scRDEN方法，这是一个强大的框架，通过将细胞中不稳定的基因表达值转换为相对稳定的基因-基因相互作用（全局特征）并提取差异表达顺序（网络特征），进而整合不同降维方法的表达特征，来推断每个分支分化方向上的重要细胞亚群和多个基因的差异表达网络。当应用于来自人类和小鼠细胞分化的五个已发表的scRNA-seq数据集时，scRDEN不仅成功捕获了具有潜在标记基因的稳定细胞亚群，测量了基因对的转录差异以识别每个分支分化方向上的等级差异表达网络。此外，在多个基因等级差异表达网络中，与转录因子/标记基因直接相关的等级表达随着它们的表达变化呈现出显著的增强和减弱趋势，多样性和聚类系数的分布呈现出非单调变化趋势，包括先增加后减少或先减少后增加的情况。这可能对应于细胞逐渐分化为稳定功能的机制。特别值得注意的是，scRDEN方法应用于大规模、多分支、双批次小鼠齿状回数据时产生了优异的结果。这种出色的性能为大规模、多批次轨迹推断以及分化和发育过程中转录机制调控的研究提供了新颖且有价值的见解。
