Fig. 12
From: recount3: summaries and queries for large-scale RNA-seq expression and splicing
Differentiating bulk and single-cell RNA-seq data. a The distribution of zeroes for different types of labeled human data. b–d PCA of human data colored by b library size; c curated labels; d predicted and curated labels. e Percent of zeroes from labeled mouse data. f, g Predicted labels from mouse data f bulk, g single-cell. The percentage of variance explained by the PC is given in the axis labels. Legends for (a-d): Bulk, manual — manually curated bulk samples; Sc, manual — manually curated single-cell samples; Sc, text mining — single-cell samples labeled by text analysis ; Small-RNA, manual — manually curated samples which were identified as small-RNA-seq; Other, manual — manually curated samples which were identified as other assay such as ribosome profiling; Unlabeled — all other samples before we differentiated bulk and single-cell samples using the percentage of zero as the predictor; Predicted Bulk — predicted bulk samples using the predictor; Predicted Sc — predicted single-cell samples using the predictor. Before doing PCA, we removed some samples which we believe are not RNA sequencing samples, which accounts for the sample number difference between (a) and (b–d)