Fig. 2
From: A statistical approach for identifying differential distributions in single-cell RNA-seq experiments
Comparison of modality in bulk versus single cells. Bar plot of the proportion of genes (or transcripts) in each dataset where the log-transformed nonzero expression measurements are best fit by a 1, 2, or 3 + mode normal mixture model (where 3+ denotes 3 or more). Modality is determined using a Bayesian information selection criterion with filtering (see “Partition estimation”). Red shades denote bulk RNA-seq datasets, and blue shades denote single-cell datasets. The number following each dataset label indicates the number of samples present (e.g., GE.50 is a bulk dataset with 50 samples). Datasets GE.50, GE.75, and GE.100 are constructed by randomly sampling 50, 75, and 100 samples from GEUVADIS [56]. Dataset LC consists of 77 normal samples from the TCGA lung adenocarcinoma study [57]. For details of the single-cell datasets, see “Methods”