[BioC] HTqPCR normalization issues - third posting

[Levi Waldron]

Dear Alessandro & Elena,

I have not used this package but have used the kinds of qPCR
normalization methods you are talking about and can offer some
comment, below.

On Mon, Nov 11, 2013 at 8:17 AM, alessandro.guffanti at genomnia.com
<alessandro.guffanti at genomnia.com> wrote:
> Now, this was our initial doubt and still persists: why a normalization
> method which should operate an average Ct value for each sample, and
> scales all Ct values according to the ratio of these mean Ct values
> across samples, should take as a reference one of the samples (the first
> in the data matrix by default) and leave its values untouched ?

I think that using average Ct value and a reference sample for HT
qRT-PCR should work as follows, if you assume 100% efficiency for all
primers:

1) for each sample, subtract the mean Ct value from the Ct value of each gene.
2) for each gene, subtract the Ct value of that gene in the reference
sample.  delta-delta-Ct for the reference sample is then zero for all
genes.

This is the method of Pfaffl 2001 (PMID 11328886) but with mean Ct
value for the sample replacing use of a housekeeping gene.  I wrote a
function a few years ago to do this, with options to specify the
efficiency of each primer or use one or more housekeeping genes, and
to specify one or more control samples (in which case the mean delta
Ct of these samples is used as reference, so that delta-delta-Ct
values for each case sample can be directly interpreted as log2(FC) ).
 My function isn't packaged or documented, but if it might be useful
to you I'd be glad to help.

> ==> *As far as scaling ranking is concerned, we noticed that even if we
> specified the reference as pseudo mean or median, apparently nothing was
> changed in term of scaling factor and reference column (always the first
> sample by default). Is it correct? Where the normalization is influenced
> by the reference?
>
> In general, which is the logic of selecting one sample as a 'reference'
> in the normalization step with these methods ?

I think it is the same logic as described by Pfaffl.  Delta-delta-Ct
values can be interpreted directly as log2(Fold Change) relative to
the reference sample.  If you have primers with efficiencies other
than 100%, then it also corrects for differences in the number of
amplification cycles for a gene between samples.

Hope this helps,
Levi

-- 
Levi Waldron
Assistant Professor of Biostatistics
Hunter College School of Urban Public Health
City University of New York
2180 3rd Ave Rm 538
New York NY 10035-4003
212-396-7747