[BioC] How to normalize Illumina bead-level data ?

[Jonas Mandel]

Hi everyone,

Does anybody know a way to work with bead-level Illumina SNP data ?

I'm interested in the correction of potential spatial bias on Illumina
SNP arrays, especially the HumanCNV370-Quad beadChip. Usually people
work with summarized data out from BeadStudio (either text files or idat
files), but here I need to process the data in a different way : the
correction on spatial bias must take place on the images of the array,
i.e on bead-level data (.txt or .TIFF files). Hence I need to first
apply my correction algorithm on the raw data, and then summarize the
data to get the X and Y signal (one value per SNP) necessary to compute
log R ratios and B allele frequencies.

I found 2 packages to work on Illumina SNP data : beadarraySNP and
crlmm. But we can only use idat files with these packages. So I use
beadarray to handle the bead-level data files (I prefer to use the .txt
files rather than the .TIFF images as the processing of .TIFF images
takes time). But I'm still searching for a way to compute summarized
data from bead-level data. I tried to reproduce the normalisation
procedure described in the white paper « Illumina's Genotyping Data
Normalization Methods » (cf. joined file), which details the procedure
described in the founding article by Piffer et al. (ref : Peiffer,
Gunderson  et al., 2006, Genome Research). I believe this normalization
procedure is the one implemented within BeadStudio.

Here is the problem : appart from any normalization or correction of
spatial bias, starting from bead-level data (either from .txt files or
from TIFF images) I fail to obtain summarized values of X and Y (or R
and Theta) identical or even close to X and Y values returned by
BeadStudio for the same array. The data I get after a « by hand »
normalization following the Illumina procedure is much more noisy and
the signal to noise ratio is much lower, when compared to the output of
beadStudio for the same array. Something go wrong but I can't identify
what. If needed I can send a file with graphs show if the comparison.

So here I need help : Does anybody know a way to work with bead-level
Illumina SNP data ? How to normalize bead-level data, i.e compute the
summarized X/Y (or R/theta) values from Illumina bead-level data ?

Any help will be more than welcome.

Best regards,

Jonas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Illumina_NormalizationWhitePaper_revD_070820.pdf
Type: application/pdf
Size: 148631 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/bioconductor/attachments/20100429/867c83bb/attachment.pdf>