[BioC] HTqPCR question

[Heidi Dvinge]

Hello Ken,

thanks for your kind words about HTqPCR, I'm glad you find it useful.

> Hello Dr. Dvinge,
> First, I would like to thank you for the HTqPCR package. It's been very
> useful for me in analysing results from a new project involving TLDA cards
> that I'm about ready to publish.
> I also have a question: do you have any advice on the best way to compare
> changes in Ct data across time points? That is, to query whether
> expression difference from Time A to Time B is different for two outcome
> classes (or whatever)? Can I simply calculate the Ct differences and read
> these data into HTqPCR, or is there a better/more elegant solution?

First of all, you don't need to calculate any Ct differences "externally",
and then reading the data into HTqPCR. Calculating Ct differences will
automatically be done within HTqPCR when you do the testing for
differential expression.

Rather than doing a simple t-test (or paired t-test depending on your
setup), it sounds like your analysis would be amenable to some of the more
sophisticated models developed in the package "limma". These are developed
for microarray data, but the same principles can be applied to some qPCR
data, and are implemented in HTqPCR in the function limmaCtData, in case
you're not already using this. The limma users guide has some excellent
user examples, and is available from within R by typing:
> limmaUsersGuide()

>From your description, it sounds like you have a factorial design, in
which case you can also analyse the interaction term, i.e. identify genes
that behave differently over time in your two outcome classes. There's an
example of this in section 8.7 of the limma guide, including instructions
on how to set up a design and contrast matrix. The difference is that
within HTqPCR the steps involving lmFit(), contrasts.fit(), eBayes() and
topTable() are all included with limmaCtData().

If this is not what you meant, then perhaps if you can provide more
details about what you've tried so far, then either myself or someone else
on the list can chime in with suggestions.

HTH
\Heidi


> (I'm finding that expression data at a given time point (pre-infection or
> 10 days post-infection) do not differ for two outcome classes, but that
> changes from the pre-infection baseline may predict outcome.)
> Thanks in advance for any insights you can provide.
> Regards,
> Ken
>
> Kenneth Witwer
> Postdoctoral Fellow
> The Johns Hopkins University School of Medicine
> Department of Molecular and Comparative Pathobiology
> 733 N. Broadway, Rm 810
> Baltimore, MD  21205
> 410-955-9770 (phone)
> 410-955-9823 (fax)
>