Identification of cell identity markers is an essential step in single-cell omics data analysis. Current marker identification strategies typically rely on cluster assignments of cells. However, cluster assignment, particularly for developmental data, is nontrivial, potentially arbitrary, and commonly relies on prior knowledge. In response, we present SEMITONES, a principled method for cluster-free marker identification. We showcase and evaluate its application for marker gene and regulatory region identification from single-cell data of the human haematopoietic system. Additionally, we illustrate its application to spatial transcriptomics data and show how SEMITONES can be used for the annotation of cells given known marker genes. Using several simulated and curated data sets, we demonstrate that SEMITONES qualitatively and quantitatively outperforms existing methods for the retrieval of cell identity markers from single-cell omics data.

中文翻译：

---

使用 SEMITONES 从单细胞组学数据中识别与簇无关的标记特征

细胞身份标记的鉴定是单细胞组学数据分析的重要步骤。当前的标记识别策略通常依赖于细胞的簇分配。然而，聚类分配，特别是对于发育数据来说，是不平凡的，可能是任意的，并且通常依赖于先验知识。作为回应，我们提出了 SEMITONES，这是一种用于无簇标记识别的原理方法。我们展示并评估其在从人类造血系统的单细胞数据识别标记基因和调控区域方面的应用。此外，我们还说明了其在空间转录组学数据中的应用，并展示了如何使用 SEMITONES 来注释已知标记基因的细胞。使用多个模拟和策划的数据集，我们证明 SEMITONES 在定性和定量上都优于从单细胞组学数据检索细胞身份标记的现有方法。