Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Bi-order multimodal integration of single-cell data

Fig. 3

Integrating single-cell RNA with protein data produced by a CITE-seq assay. ac UMAPs of 30,672 human bone marrow cells based on abundance of the 25 surface proteins (a), RNA expression levels of 3000 highly variable gene (b), and RNA expression levels of the 25 protein-coding genes (c). Labels and dots are colored synchronously by cell type information from the original study. The ARI values are labeled in each panel. d, e UMAPs of the protein (d) and the RNA (e) expression data in the co-embedding generated by bindSC. Each dot in the boxplot denotes one cell type. f Label transfer accuracy of bindSC, Seurat v3.0, LIGER, and Harmony. Each dot in the boxplot denotes one cell type. g Improvement in accuracy of imputed protein level. Each dot represents a protein. X-axis is the Pearson correlation between the ground truth protein level and the RNA level of its coding gene. Y-axis is the Pearson correlation between the ground truth protein level and bindSC imputed protein level. h Comparison of the epitope abundance of CD19, CD14, and CD11c (x-axes) with the RNA expression levels of their coding genes (i.e., CD19, CD14, and ITGAX; y-axes; first row) and with the bindSC imputed protein levels (y-axes; second row). i Gene-protein network inferred from Pearson correlation between genes and bindSC inferred protein levels. A cutoff of 0.55 is used and top five highly correlated genes of each protein are kept

Back to article page