Beginner question: box-plot and PCA in RNA-seq

Question

[scchess]

This is a beginner question.

I have drawn up a box-plot and PCA plot for my RNA-seq experiment before and after normalization. However, I don't understand how to interpret the plots. For example, in the PCA plot, are the biological replicates expected to cluster by the first principal component? What about cluster on the second principal component? If I see my samples are clustered in a box-plot, does that mean my normalization has done the job?

What does a perfect normalization would look like on a box-plot and PCA plot? Any online resources?

Answer 1

[Devon Ryan]

I'm not entirely sure what you're plotting in your box plot (normalized counts?), so I won't comment on that.

When you plot the PCA results you don't necessarily expect the biological replicates to cluster by group (either in PC1 or PC2), that depends entirely on the effect size and the underlying biological noise. The general use of a PCA plot is to ensure that you don't have any unknown batch effects causing outliers or really weird clustering of samples.

[scchess]

Thanks for the explanation. I made a box-plot and PCA both before and after normalization. It was done on the raw and normalized counts. Does that mean technical replicates are expected to cluster by the first or second component in a PCA plot? I'm trying to understand how to read the plots.

[Devon Ryan]

If you have technical replicates then they should always cluster together after normalization. If not then something went wrong (not in normalization, but in making one of the replicates).