[BioC] TilingArray Normalization

Wolfgang Huber huber at ebi.ac.uk
Sat Nov 29 12:13:07 CET 2008

Hi Anjan,

it is quite crucial for the "normalizeByReference" method that most of
the "background" probes really have untranscribed sequence - does that
seem to be case when you plot their intensities? How does the histogram
of their intensities look like compared to the known transcribed regions?

Also, the goal of that method is not to reduce noise, but to increase
the signal-to-noise ratio. That is a subtle, but important difference.
Have a look at Figure 5 of
http://bioinformatics.oxfordjournals.org/cgi/reprint/22/16/1963.pdf

Can you be more precise than "The "normalized" signal after the
tilingArray normalization seem to be *noisier* than the non-normalized
input." How do you measure this? Do you have a plot to support this?
Perhaps Fig.5 and the quantity $\Delta\mu / \sigma$ in that paper might
be useful for that purpose.

Thank you and Best wishes
Wolfgang

----------------------------------------------------
Wolfgang Huber, EMBL-EBI, http://www.ebi.ac.uk/huber

Anjan Purkayastha ha scritto:
> Hi,
> I am trying to use the tilingArray normalization package for the
> following array:
> total number of probes: 6490. (these are a set of non-overlapping,
> contiguous 60mer probes that span the vaccinia virus genome-~200kb)
> total number of perfect matches: 6490.
> total number of background probes: 1275.( the vaccinia genome is very
> densely packed so there are few untranscribed regions).
> The above mentioned array is being use to assay the transcription of the
> viral genes over an infection time-course.
>
> Issues:
> The probe intensities from DNA-hybs used for normalization are on
> average as strong as the non-normalized probe intensities from
> transcription assays.
> The "normalized" signal after the tilingArray normalization seem to be
> *noisier* than the non-normalized input.
>
> I know there may not be enough material in this email for the Gurus to
> I'd be happy to provide more details- genomic map of normalized and
> non-normalized signal etc, but there seems to be a size limit to
> attachments for this forum.
> Please let me know if you have any specific questions regarding the
> experiment and/or the data.
>
> All advice/questions will be appreciated.
> Anjan
>