RNA-Seq Data Analysis    ◾    187

share a dispersion. Three types of dispersions are calculated: a common estimate across

all genes, mean-variance trend dispersion using genes’ similar abundance, and gene-wise

dispersion (tagwise dispersion).

The “estimateDisp(DGEList, design)” function estimates the common, trended, and

tagwise negative binomial dispersions by using weighted likelihood empirical Bayes algo-

rithm [33]. This function requires a DGEList object with normalized counts and a design

matrix.

yNorm <- estimateDisp(yNorm, design)

names(yNorm)

The three dispersions will be estimated and stored in the DGEList object (y) as shown in

Figure 5.12. You can display the common, trended, and tagwise dispersions by using the

following (Figure 5.13):

yNorm$common.dispersion

head(yNorm$trended.dispersion)

head(yNorm$tagwise.dispersion)

FIGURE 5.12  DGEList objects slot including the estimated dispersions.

FIGURE 5.11  DGEList object after computing the normalization factor.