Figure 1
From: Statistical measures of transcriptional diversity capture genomic heterogeneity of cancer
Assessment of different transcriptional diversity metrics in simulated datasets. A) Simulated gene expression profiles generated using a hierarchical model to independently control within sample (σ g ) and between samples (σ p ) transcriptional variation and the number of latent subgroups. Each profile consists of 50 genes (rows) and 40 samples (columns). Profiles for 1, 2, 4 and 40 latent subgroups are shown for low (σ p /σ g  = 0.5/1.5) and high (σ p /σ g  = 0.5/0.5) relative between-to-within sample variation. B) Transcriptional diversity within the simulated profiles assessed using the mean pairwise Pearson distance, the mean pairwise cosine distance, or the mean dispersion distance. The boxplots represent the distributions of these metrics obtained from 500 independent simulations of each dataset (blue: low σ p /σ g , green high σ p /σ g ). C) Same metrics as in B assessed in a two latent subgroup dataset of 40 samples with increasing proportion of the smaller subgroup. Boxplots represent distribution over 500 independent simulations.