Differential expression in qPCR Data after filtering

Question

[Juan Cordero]

Dear community,

I have qPCR data samples with a lot of noise and a general low amplification. Many Ct values are set therefore to 40 (or NA, I'm using R package HTqPCR). My question is whether I should filter out "genes where the results are of low quality" before performing DE analysis, and how I can trust the p-values. Are results not going to be misleading if I significant fraction of the genes is removed from the analysis?

Thanks in advance

Answer 1

[Devon Ryan]

You can go ahead and discard the Ct ~40 genes, you don't have the ability to measure any changes in them to begin with. The results won't be misleading if you remove them.

[Anima Mundi]

Hi Devon, I see your point. However, what if a gene becomes measurable (or unmeasurable) only after the treatment? Also, if he ends up removing, say, 60% of the studied genes, the overall conclusions might be misleading (e.g. he might think his treatment favors up-regulation of a certain class of transcription factors, that were instead simply comparatively more easy to measure).

[Devon Ryan]

If only one group has such a high Ct then that's not a problem. Filter by the average or median Ct across groups.

Answer 2

[Anima Mundi]

Yes, they are going to be misleading. I would rather try to optimize your wetlab workflow (e.g. changing PCR primers).