[BioC] limma: print-tip loess and empty spots

[Gordon K Smyth]

Dear Adrian,

I assume that you're already read the limma User's Guide advice:

"Print-tip loess is also unreliable for small arrays with less than, say, 150 spots per print-tip
group. Even larger arrays may have particular print-tip groups which are too small for printtip
loess normalization if the number of spots with non-missing M-values is small for one or more of
the print-tip groups. In these cases one should either use global "loess" normalization or else
use robust spline normalization"

There are however special considerations for multispecies arrays, see

Gilad, Y., Oshlack, A., Smyth, G. K., Speed, T. P., and White, K. P. (2006). Expression profiling
in primates reveals a rapid evolution of human transcription factors. Nature 440, 242-245.

Oshlack, A., Smyth, G. K., and Gilad, Y. (2007). Using DNA microarrays to study gene expression in
closely related species. Bioinformatics. (Published online 23 March 2007).

and perhaps

Oshlack, A., Emslie, D., Corcoran, L., and Smyth, G. K. (2007). Normalization of boutique
two-color microarrays with a high proportion of differentially expressed probes. Genome Biology 8,
R2.

Best wishes
Gordon

------------- original message ----------------
Adrian Steward adrian.steward0405 at gmail.com
Mon May 21 19:52:56 CEST 2007

Hi all,

I am using the limma package to analyze a multi-species cDNA array, 2-colour
reference design.  The problem is that because it is a multi-species (and
tissue) array, and I am querying only 1 tissue, only 1/3 of the 15,000 spots
appear to correspond to cDNA in my samples, and the number of spots that
actually get tested is around 3,500.  These spots are rather randomly
located across the slides because of many libraries used in construction of
the array.

Before I get too far into my analysis, I read in the limma guide that
print-tip loess within-slide normalization is not always a good choice for
data with 'small' print tip groups.  I am assuming that a global loess
normalization is the more appropriate approach in my case.

Is my assumption reasonable?

With thanks

Adrian M.


PS - I'm running R 2.3.1, limma 2.7.3, and limma GUI version 1.8.1