Explore gene expression variability: PCA

[lcolladotor]

This can be done in different ways.

• Plot the top principal components against each other and color them by some covariates we have.
• Plot top PCS one at a time using boxplots or scatterplots against covariates of interest

For the first situation, see the first plot from http://research.libd.org/SPEAQeasy-example/de_analysis_speaqeasy.html#15_Explore_and_visualize_gene_expression

For the second, see the the boxplots (figs 2 and 3 at http://research.libd.org/SPEAQeasy-example/de_analysis_speaqeasy.html#15_Explore_and_visualize_gene_expression).

We have several variables of interest like:

• study type (since each experiment was done in a different mouse facility)
• adult vs pup
• exposed vs non exposed
• pup-dame families (we have that info for just some mice)

[lcolladotor]

See http://research.libd.org/recount-brain/example_multistudy/recount_brain_multistudy.html#run-pca for some code for plotting several PCs (1 vs 2, 3 vs 4, 5 vs 6 and 7 vs 8)

[lcolladotor]

As noted in #3, we also want to compare the PCs vs SPEAQeasy output summary metrics like the mitochondrial chr mapping rate, the total gene assigned rate, the rRNA rate, the ERCC error sum, etc.