Significant Expression Difference

Question

[Adrian Pelin]

Hello,

I am doing some preliminary RNA-Seq and I have 2 libraries, 1 for condition A and one for condition B. I know how to use bowtie and cufflinks to estimate FPKM values for genes of interest, but how can I get a list of genes which have a significant expression difference so that I may test them with qRT-PCR?

Thank you.

Answer 1

[Devon Ryan]

1. Without replicates you will never get meaningful p-values. You might consider just sorting things by fold-change.
2. If you've already run cufflinks, just use cuffdiff (it'll give you p-values, but they're not very meaningful without replicates).

[Adrian Pelin]

My thinking is that I would use q-RTPCR to validate the data after finding potential candidates. Do you think the approach is reasonable?

[Devon Ryan]

Yeah, though keep in mind that your validation rate might be really low. At least if you're using Taqman assays, it usually ends up being cheaper to just sequence more samples (then you end up wasting fewer assays).

[Nicolas Rosewick]

and wasting time. just sequence one or two additional samples for each group. If you only look for expression, you don't have to go deep, so you can multiplex all your samples in one lane.

[Devon Ryan]

Very true, though the trick is to have someone else do the qPCRs for you. Wasting time only sucks when it's your time being wasted :p

[Adrian Pelin]

It's not my project, I am just in charge of the bioinfo part:) They will be using SYBR Green super mix, and they are only interested in a few genes, so hopefully we don't waste too much money.

We did 50bp single end sequencing, I think that's enough since we aren't trying to reconstruct transcriptomes here...

[Nicolas Rosewick]

You can try DESeq, there is a "without replicate" mode