Fig. 5
From: Combining single-cell ATAC and RNA sequencing for supervised cell annotation
Sensitivity of ATAC-derived improvements to ground truth cell type granularity.. The addition of ATAC features only leads to significant improvements in classification performance for the more granular WNN-L2 cell subtype labels. Boxplots show macro F1, precision and recall scores per class using scVI embeddings with an SVM classifier. Results are stratified by ground truth containing cell subtypes (WNNL2) or major cell types only (WNNL1). Boxes indicate the median, lower and upper quartile over 10 bootstrap out-of-bag test samples for macro-average metrics. Whiskers indicate 1.5 times the interquartile range and dots indicate outliers beyond the whiskers. Wilcoxon-signed rank test with Bonferroni correction was used to compare RNA-alone to RNA and ATAC models (* = p < 0.05)