search

davismcc / scaterPaperExtras

Discussion board for modifications to the scater paper
0 stars 0 forks source link

Page 6: Data normalisation and batch correction: Paragraph 4 #21

Closed LTLA closed 8 years ago

LTLA commented 8 years ago

Applying removeBatchEffect from limma yields normalised data for which PCs are no longer correlated with C1 machine (see Case Study in Supplementary Material), but the patient effect is not the primary driver of variation amongst the PCs (Figure 4c).

removeBatchEffect depends on a known batch effect, so why is it in the paragraph with unknown effects? Also, it doesn't seem to work properly in 4C (or at least, not as well as 4B), so why not put your best foot forward and just leave out the rest?

Perhaps adding an extra sentence at the end to conclude:

This type of correction is crucial to mitigate the effect of uninteresting technical factors on the biological interpretation of any variation or population structure.

davismcc commented 8 years ago

Yes, good point about removeBatchEffect. Just because it can be used doesn't mean it should. Removing this would also simplify the figure, which is needed. I'll drop that section from the text and revise the figure. That can make the figure a single-column with three panels, which will be much more readable.