Fig. 4

Model-based smoothing and cis-regulatory element inferring. By smoothing and generating mini-bulk omics profiles summing up neighborhood cells, we can infer the gene regulatory regions by calculating the correlation between transcriptome and chromatin openness. A higher correlation indicates a likely regulatory relationship between genes and peaks. A (Left) Smoothing decreased the sparsity of scRNA-seq and scATAC-seq data. (Middle) Smoothing increased the correlation between gene expression and its TSS regions openness compared with non-smoothed and true pair derived mini-bulk data. (Right) This trend is emphasized when showing the Spearman correlation coefficient differences between smoothed and non-smoothed mini-bulk data. B Example showing FGF14 and its TSS region peaks correlation in non-smoothed and smoothed data. C Heatmap showing the mean of correlation level between gene-peak pairs with different distances in all smoothed and non-smoothed datasets. D Genome track of LEF1 and its highly correlated peaks. Left shows the genome tracks of ATAC-seq data, right violin plots show the gene expression level