Skip to main content
. 2013 Jul 2;8(7):e67019. doi: 10.1371/journal.pone.0067019

Figure 2. Sampling and technical variance is distinct from within-condition variance.

Figure 2

A comparison of within-condition gene expression-difference to the median expression level for three different experiments. The left panel shows the sampling variance for comparison and the three experiments are shown in subsequent panels. The L-K RNA-Seq data set compares gene expression in two liver and two kidney samples. The Bacillus cereus RNA-Seq data set for samples grown at neutral pH and two samples 20 minutes after shift to grown at low pH. Meta-RNA-Seq data is for microbial gene expression analysis of four clinical vaginal samples from two women with a healthy microbiota and two women with a microbiota indicative of bacterial vaginosis. The RNA-Seq experiments in the L-K and B. cereus datasets were from controlled conditions with identical gene content per condition and show that the vast majority of highly-expressed genes have small within-condition estimates of Inline graphic, and that estimate only becomes imprecise as Inline graphic becomes very small. The Meta-RNA-Seq panel shows that when within-condition variance is high, there is no relationship between the expression level and the within-condition variance. Note that base-2 logarithms were used throughout.