[RNASeq] Excluding samples that cluster differently in PCA plot

Question

[Texx]

Would you exclude the samples that "clearly" cluster differently in a PCA plot after the analysis with DESeq2 (see below)? Particularly, I'm wondering whether I should exclude samples LAC_8_7 and LAC_8_9 from further data analysis solely based on these differences?

Those two samples were greatly contaminated with rRNA (depletion failed) but all the rRNA counts were removed from the count table before the analysis.

Answer 1

[LChart]

You either need to remove them, treat them as a different batch, or include a variable that is a surrogate for that batch. You might try to remove raw %rRNA (or total rRNA or log-total-rRNA) as a quantitative covariate by incorporating it into batch correction methods, i.e. via covar_mod in combat-seq.