[BioC] Creating a CDFenv without cdf file

[lgautier@altern.org]

> Hi,
> I have many cel files from a custom chip, but I do not have the cdf
file. I do have the chip design information and
> and the seqence information but they are in seperate text files. I
looked in the altcdfenvs package and it looks
> as if I could hijack another env and change the parameters as needed.
Has anyone done this?

Well, building you cdfenv from scratch would be easier I think.

The package altcdfenvs proposes a class to wrap cdfenvs and convenience
methods to slice, dice, and do basic sanity checks on that beast.
Once I hoped that this class, or actually a class, would make it to the
rest of "affy" and "makecdfenv", but apparently it did not.

You can use the class when building your thing then extract the
environment from its slot when done, or go directly with an environment.

> Is there a tutorial
> or Vignette I overlooked?

No tutorial for this exact sequence of steps, but looking at the code that
makes cdfenvs from the results of matchprobes in altcdfenvs should help
you to write your own.
In short, you will want something like:
- read through your 'design files' and get for each probe its
location (X/Y coordinates) and the probeset it belongs to.
   |
   -- resolve each X/Y pair of coordinates to an 'index'
(that is the row index for the probe level data in the AffyBatch object)
 using the function xy2indice (note that you will need the geometry for
your chip -number of rows/number of cols-... if you do not have it, try
to take the square root of the number of probe-level values in your cell
file: Affymetrix chips are often squares).
   |
   -- populate your cdfenv environment with keys that are probeset
identifiers ("12345_at"-like strings) and values that are two-columns
matrices (first column being PM, second column being MM... if nothing
is known about MMs and no plan to use them set them all to NA, otherwise
there is a dangerous empirical rule saying that MMs are next to their
respective PM).


> Is there another tool for building cdf files?

None that I know of... may be the interface to the affymetrix
readers/parsers ?



Hoping this helps,



L.


> thanks
> Nicholas
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
>