DESeq2 model

RNA-Seq with Bioconductor in R

Mary Piper

Bioinformatics Consultant and Trainer

DESeq2 model

DE theory

DESeq2 model - mean-variance relationship

# Syntax for apply()
apply(data, rows/columns, function_to_apply)

# Calculating mean for each gene (each row)
mean_counts <- apply(wt_rawcounts[, 1:3], 1, mean)

# Calculating variance for each gene (each row)
variance_counts <- apply(wt_rawcounts[, 1:3], 1, var)

DESeq2 model - dispersion

Plotting relationship between mean and variance:

# Creating data frame with mean and variance for every gene
df <- data.frame(mean_counts, variance_counts)

ggplot(df) +
        geom_point(aes(x=mean_counts, y=variance_counts)) + 
        scale_y_log10() +
        scale_x_log10() +
        xlab("Mean counts per gene") +
        ylab("Variance per gene")

DESeq2 model - dispersion

mean-variance relationship

DESeq2 model - dispersion

$Var$: variance

$\mu$: mean

$\alpha$: dispersion

Dispersion formula: $Var = \mu + \alpha * \mu^{2}$

Relationship between mean, variance and dispersion:

$$\uparrow variance \Rightarrow \uparrow dispersion$$

$$\uparrow mean \Rightarrow \downarrow dispersion$$

DESeq2 model - dispersion

# Plot dispersion estimates
plotDispEsts(dds_wt)

Dispersion plot

DESeq2 model - dispersion

bad_dispersion_plots

Let's practice!

RNA-Seq with Bioconductor in R