[BioC] Error in function (classes, fdef, mtable): unable to find an inherited method for function "indexProbes", for signature "exprSet", "character"

Sean Davis sdavis2 at mail.nih.gov
Sun Mar 23 23:22:27 CET 2008

On Sun, Mar 23, 2008 at 5:33 PM, Suprabhath Reddy <sgajjala at emich.edu> wrote:
> Sean Davis,
> I am thankful for your Quick reply to my posting.
> Yes, Illumina data and Affy data are different. But, I am in a situation
> where I need to compare/correlate the expression values from two experiments
> which are totally different. In the first experiment, the RNA expression
> values are collected by using Affymetrix chips and in the second experiment
> the RNA expression values are collected by using Illumina chips.
> The raw data is not provided by the first experiement authors in the GEO
> site. The first experiment authors have just provided normalized data(Rank
> Invariant Normalization from BeadStudio Software). And the second experiment
> authors have provided both raw data and normalized data(Quantile
> Normalization followed by Median Normalization).
> The interesting thing is that both the authors have collected the expression
> values for almost same set of genes , but in a different tissue (cortex and
> Lymphoblastoid cell lines)in their experiments.
> So, I am trying to normalize the raw data provided by second experiment
> authors by using the rank.invariantset()  method  in R. I believe that if I
> can normalize(Rabk invariant) the raw data(from second experiment) , then I
> can compare the expression values because both the datasets have been
> normalized using the same metric.

Hi again, Suprabhath.

For future reference, feel free to reply back to the list.  It lets
everyone learn from the exchange.

If what you are saying is that you have data from one tissue type done
on affy arrays and data from another tissue type done on illumina,
then you almost certainly cannot compare the raw OR normalized values
to each other.  While the same genes might have been measured, the
probes and the technologies are not comparable.  Even if you had data
on a single platform (eg., Illumina), the data may still not be
directly comparable.

I may be misunderstanding your intentions.  If so, it might be a good
idea to clarify what you are trying to do (besides invariant set
normalization), as what I think you are describing is not possible, at
least in the simple way you are describing.


More information about the Bioconductor mailing list