The CBMC cell types annotated by Seurat can be found at the following website: https://satijalab
The CBMC cell types annotated by Seurat can be found at the following website: https://satijalab.org/seurat/archive/v3.2/multimodal_vignette.html. The cell types of the HBMC dataset annotated by Seurat are derived from the analysis on this webpage: https://satijalab.org/seurat/articles/weighted_nearest_neighbor_analysis.html. The external data sets used to test operational efficiency come from the 10x Genomics website, and they can be obtained from our GitHub repository (https://github.com/studentiz/CITEMO/tree/main/Data/10x). Supplementary material Supplemental data for this article can be accessed here.. framework, we apply CITEMO to analyse the cell subtypes of Cord Blood Mononuclear Cells (CBMC) samples. Results show that the CITEMO framework can comprehensively analyse single-cell multimodal samples and accurately identify cell subtypes. Besides, we find some specific immune cells that co-express multiple ADT markers. To better describe the co-expression phenomenon, we introduce the co-expression entropy to measure the heterogeneous distribution of the ADT combinations. To further validate the robustness of the CITEMO framework, we analyse Human Bone Marrow Cell (HBMC) samples and identify different states of the same cell type. CITEMO has an excellent performance in identifying cell subtypes and states for multimodal omics data. We suggest that the flexible design idea of CITEMO can be an inspiration for other single-cell multimodal tasks. The TH588 hydrochloride complete source code and dataset of the CITEMO framework can be obtained from https://github.com/studentiz/CITEMO. (Fig. 2D) [27]. Since the abundance of mouse-associated ADT is not measured in CBMC samples, the mouse cells could not be identified by ADT information alone. At the same time, relying only on genetic information may also miss some important cell subtypes. For example, only a population of NK cells are found with the transcriptome analysis, while the ADT analysis successfully identifies three subtypes of NK cells, including CD56bright NK [38C40], CD8- NK [41,42] and CD8+?NK [42C44] (Fig. 2E and Supplementary Figure 1A). These three subtypes are also supported SLC2A4 by the previous studies [45]. The different recognition ability of cell subtypes for transcriptome and ADT analysis is caused by the difference in PCA analysis. We use PCA to convert the expression of genes and ADT into PCs. The distributions of PCs of cell subtypes in transcriptome and ADT modalities are specific, which implies that PCs represent specific information about cell subtypes (Figs. 3D&3E). For example, PC1 in the transcriptome modality is closely related to mouse genes (Figs. 3G&3?J), while the distribution of PC1 in ADT mode is consistent with the distribution of CD4?+?T cells (Figs. 3H&3?K). Integrating the low-dimensional representation of the transcriptome and ADT data can obtain a multimodal omics representation, giving a more comprehensive characterization of the heterogeneity of cells. To avoid the error introduced by different algorithms, we keep using the PCA algorithm to integrate multimodal omics. We apply the elbow method to select the first 15 PCs as the low-dimensional representation of multimodal omics (Fig. 3C). Then, the cell clustering process is performed for the low-dimensional representation of multimodal omics, followed by the manual annotation of cell types. The results given by the CITEMO multimodal omics cover all cell subtypes identified by CITEMO transcriptome and CITEMO ADT separately TH588 hydrochloride (Fig. 2C, Supplementary Figures 1A&2C). CITEMO multimodal omics successfully identifies 3T3 and 4T1 mouse cells, as given by CITEMO transcriptome (Fig. 2D). TH588 hydrochloride Moreover, CITEMO multimodal omics also successfully identifies three subtypes of NK cells (Fig. 2E and Supplementary Figure 1A), as given by CITEMO ADT. Another noteworthy finding is that CITEMO multimodal omics identified CD16+?CD45RA+ monocyte which was annotated as NK cells TH588 hydrochloride by a previous study (Fig. 2F) [46,47]. CD16+?CD45RA+ monocytes are similar to some NK cells, in that they all express CD16 (Fig. 2F and TH588 hydrochloride Supplementary Figure 1A). This may be the reason why it was identified as NK cells by previous methods [3]. However, CD16+?CD45RA+ expresses CD14 and CD11c, which are markers of monocytes (Fig. 2F and Supplementary Figure 1A) [48]. Therefore, we believe that it is not an NK cell but a special type of monocyte. Alternatively, CD16+?CD45RA+ monocytes may be the activated monocytes due to their higher CD45RA expression than other types of monocytes (Fig. 2F). This implies that CITEMO framework can detect the cell states. We further compare the results given by CITEMO multimodal omics with previous studies [3]. CITEMO multimodal omics identifies naive CD4 T cells and memory CD4 T cells from the CD4 T cells according to the abundance of CD45RA ADT.