Pre-process features

In the video exercise, you saw that the sample distributions for the doxorubicin study were extremely right-skewed. Thus, the first step you need to take is to pre-process the features: log-transform, normalize, and filter.

This exercise is part of the course

Differential Expression Analysis with limma in R

View Course

Exercise instructions

The ExpressionSet object eset_raw with the raw data has been loaded in your workspace. The limma package is loaded.

Log transform the measurements. Use plotDensities to visualize. Label the samples by their genotype.
Quantile normalize the measurements with normalizeBetweenArrays and re-visualize.
Use rowMeans to determine which genes have a mean expression level greater than 0.
Filter the genes (i.e. rows) with the logical vector keep and re-visualize.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Create a new ExpressionSet to store the processed data
eset <- eset_raw
# Log transform
exprs(eset) <- ___(exprs(eset))
___(eset,  group = pData(eset)[___], legend = "topright")
# Quantile normalize
exprs(eset) <- ___(exprs(eset))
___(eset,  group = pData(eset)[___], legend = "topright")
# Determine the genes with mean expression level greater than 0
keep <- ___(exprs(eset)) > ___
sum(keep)
# Filter the genes
eset <- eset[___]
___(eset, group = pData(eset)[___], legend = "topright")

Edit and Run Code