Skip to main content
Fig. 4 | Genome Biology

Fig. 4

From: scAI: an unsupervised approach for the integrative analysis of parallel single-cell transcriptomic and epigenomic profiles

Fig. 4

Revealing cellular heterogeneity and regulatory links of dexamethasone-treated A549 cells. a Heatmap of the cell loading matrix H obtained by scAI. Cells are ordered and divided into early, transition, and late stages based on hierarchical clustering. The bar at the bottom indicates the collection time of each cell. b Genes are ranked in each factor based on gene scores calculated from gene loading matrix, in which the known markers are indicated. c Loci are also ranked based on locus scores from locus loading matrix, in which the motifs and the corresponding logo of some TFs of the known marker genes are indicated. The binding TFs of the known marker genes and the chromosome loci of these motifs were found from hTFtarget database. d Visualization of cells by VscAI. Known marker genes (left panel) and motifs related with these marker genes (right panel) were projected onto the same low-dimensional space. The same motifs such as SMAD3 and NR3C1 are shown in two opposite positions, as they are enriched in different loci. These loci were located within 10 kb of marker genes’ regulatory regions, which were extracted from the database (http://bioinfo.life.hust.edu.cn/hTFtarget/) in lung tissue. Here, we visualized the motifs instead of individual loci for easier understanding. e The fold enrichment (FE) values of inferred regulatory links of the known markers, which were validated by the hTFtarget database. f Inferred regulatory links of gene ABHD12 for each factor and the epigenome browser visualization of DNase-seq data and NR3C1 ChIP-seq data derived from chromatin regions near TSS of ABHD12. The red stem represents the TSS region of the gene, and the black stem represents each locus. Five regulators which correspond to the inferred regulatory links were indicated. The gray region shows the distinct regulatory links (regulated by NR3C1) between 0 h and 1 and 3 h. g The pseudotemporal chromatin accessibility trajectory was inferred with the aggregated scATAC-seq data. Cells were visualized in the first two diffusion components (DCs). The gray line is the fitted principal curve. Bottom: the percentages of cells at the three time points during the inferred pseudotime, which was divided into 10 bins. h Inferred pseudotime of three key genes. The black line indicates the fitted expression levels using cubic splines. i Left: “Rolling wave” plot shows the normalized smoothed accessibility data for the pseudotime-dependent accessible loci clustered into two groups. Middle: the normalized smoothed gene expression data for the pseudotime-dependent genes along the inferred accessibility trajectory using the aggregated scATAC-seq data. Loci and genes are ordered based on the onset of activation. Right: the corresponding gene dynamics along the cellular trajectory inferred only using scRNA-seq data

Back to article page