Institute of Cytology, Russian Academy of Sciences, St. Petersburg, Russia.
FEBS J. 2020 Oct;287(20):4427-4439. doi: 10.1111/febs.15257. Epub 2020 Mar 5.
The transcriptome consists of various gene modules that can be mutually dependent, and ignoring these dependencies may lead to misinterpretation. The most important problem is module dependence on cell-cycle activity. Using meta-analysis of over 30 000 single-cell transcriptomes, we show gene module dependencies on cell-cycle signature, which can be consistently observed in various normal and cancer cells. Transcript levels of receptors, plasma membrane, and differentiation-related genes are negatively regressed on cell-cycle signature. Pluripotency, stress response, DNA repair, chromatin remodeling, proteasomal protein degradation, protein network connectivity, and unicellular evolutionary origin are regressed positively. These effects cannot be explained by partial overlap of corresponding gene sets because they remain if the overlapped genes were removed. We propose a visual analysis of gene module-specific regression lines as complement to an uncurated enrichment analysis. The different lines for a same gene module indicate different cell conditions. The approach is tested on several problems (polyploidy, pluripotency, cancer, phylostratigraphy). Intriguingly, we found variation in cell-cycle activity, which is independent of cell progression through the cycle. The upregulation of G2/M checkpoint genes with downregulation of G2/M transition and cytokinesis is revealed in polyploid cells. A temporal increase in cell-cycle activity at transition from pluripotent to more differentiated state is found in human embryonic stem cells. The upregulation of unicellular interactome cluster in human cancers is shown in single cells with control for cell-cycle activity. The greater scatter around regression line in cancer cells suggests greater heterogeneity caused by deviation from a line of normal cells.
转录组由各种相互依赖的基因模块组成,如果忽略这些依赖性,可能会导致误解。最重要的问题是模块对细胞周期活动的依赖性。通过对超过 30000 个单细胞转录组的元分析,我们展示了基因模块对细胞周期特征的依赖性,这种依赖性在各种正常和癌细胞中都可以观察到。受体、质膜和分化相关基因的转录水平与细胞周期特征呈负相关。多能性、应激反应、DNA 修复、染色质重塑、蛋白酶体蛋白降解、蛋白质网络连接性和单细胞进化起源呈正相关。这些效应不能用相应基因集的部分重叠来解释,因为即使去除重叠基因,这些效应仍然存在。我们提出了一种基因模块特异性回归线的可视化分析方法,作为无注释富集分析的补充。对于相同的基因模块,不同的回归线表示不同的细胞状态。该方法在几个问题(多倍体、多能性、癌症、系统发生学)上进行了测试。有趣的是,我们发现了与细胞周期进程无关的细胞周期活性变化。在多倍体细胞中,G2/M 检查点基因的上调伴随着 G2/M 转换和胞质分裂的下调。在人类胚胎干细胞中,从多能状态向更分化状态过渡时,细胞周期活性会暂时增加。在单细胞中,通过控制细胞周期活性,显示出人类癌症中单细胞相互作用簇的上调。癌细胞中回归线周围的离散度更大,表明由于偏离正常细胞的线性而导致更大的异质性。