Journal: Nature Communications

Article Title: Intrinsic bias estimation for improved analysis of bulk and single-cell chromatin accessibility profiles using SELMA

doi: 10.1038/s41467-022-33194-z

Figure Lengend Snippet: Visualization of intrinsic cleavage bias effect in different cell clusters derived from scATAC-seq data for different biological samples and different experimental platforms: human hematopoietic cells ( a – c ), mixed human cell lines ( d – f ), mouse primitive gut tube ( g – i ), and 10× Single-Cell Multiome data for mouse embryonic brain ( j – l ), human peripheral blood mononuclear cells (PBMC) ( m – o ), and human lymph node ( p – r ). a , d, g , j , m , p UMAP visualization where cells are colored by cell type/labels/clusters. b , e , h , k , n , q Same UMAP visualization but cells are colored by cell bias score (CBS). c , f , i , l, o , r CBS distributions of cells from different cell types/batches/clusters. Boxes are colored by cell clusters using the same color palette as the first column. The centerline, bounds of box, top line, and bottom line of the boxplots represent the median, 25th to 75th percentile range, 25th percentile – 1.5 × interquartile range (IQR), and 75th percentile + 1.5 × IQR, respectively. The cell numbers for all boxplots are listed in Supplementary Dataset .

Article Snippet: The 10x Single-Cell Multiome datasets are downloaded from 10x Genomics website ( https://www.10xgenomics.com/resources/datasets/fresh-embryonic-e-18-mouse-brain-5-k-1-standard-2-0-0 , https://www.10xgenomics.com/resources/datasets/pbmc-from-a-healthy-donor-no-cell-sorting-10-k-1-standard-2-0-0 , https://www.10xgenomics.com/resources/datasets/fresh-frozen-lymph-node-with-b-cell-lymphoma-14-k-sorted-nuclei-1-standard-2-0-0 ).

Techniques: Derivative Assay

Journal: Nature Communications

Article Title: Intrinsic bias estimation for improved analysis of bulk and single-cell chromatin accessibility profiles using SELMA

doi: 10.1038/s41467-022-33194-z

Figure Lengend Snippet: Comparisons of cell clustering accuracy before and after considering the peak bias score (PBS) on scATAC-seq data for human hematopoietic cells ( a ), mixed human cell lines ( b ), mouse primitive gut tube ( c ), and 10× Single-Cell Multiome data for mouse embryonic brain ( d ), human PBMC ( e ), and human lymph node ( f ). K-means clustering was performed after PCA dimensionality reduction. Blue, using all ATAC-seq peaks; orange, after removing 1–50% of peaks with the highest peak bias score (PBS). For each percentage of peaks retained from 50% to 99% with a 1% increment, 100 randomly sampled subsets of peaks were used as the background for determining the relative ranks of all peaks or retained peaks. The relative ranks of the adjusted Rand index (ARI), defined as the rank relative to the 100 randomly sampled sub-datasets, for the 50 cases from 50% to 99%, were plotted ( n = 50 for each boxplot). P values were calculated by the one-sided Wilcoxon signed-rank test. The centerline, bounds of box, top line, and bottom line of the boxplots represent the median, 25th to 75th percentile range, 25th percentile – 1.5 × interquartile range (IQR), and 75th percentile + 1.5 × IQR, respectively. g Schematic of SELMA single-cell peak bias correction model. Peaks are weighted and adjusted based on PBS percentile using an empirically determined weight function. h – m Comparisons of the accuracy (measured by ARI) of single-cell clustering generated using different existing tools on scATAC-seq data with (orange) or without (blue) SELMA single-cell peak bias correction. Each panel represents the result for a scATAC-seq sample from a different biological system or experimental platform: human hematopoietic cells ( h ), mixed human cell lines ( i ), mouse gut tube ( j ), mouse embryonic brain ( k ), human PBMC ( l ) and human lymph node ( m ). Each data point at the center of the error bar represents the average (mean) ARI generated from 100 runs with different random seeds using the method labeled on the x-axis. The error bar represents the standard deviation ( n = 100).

Article Snippet: The 10x Single-Cell Multiome datasets are downloaded from 10x Genomics website ( https://www.10xgenomics.com/resources/datasets/fresh-embryonic-e-18-mouse-brain-5-k-1-standard-2-0-0 , https://www.10xgenomics.com/resources/datasets/pbmc-from-a-healthy-donor-no-cell-sorting-10-k-1-standard-2-0-0 , https://www.10xgenomics.com/resources/datasets/fresh-frozen-lymph-node-with-b-cell-lymphoma-14-k-sorted-nuclei-1-standard-2-0-0 ).

Techniques: Generated, Labeling, Standard Deviation