Could someone suggest a procedure for performing scale normalization
with the vsn method on an marrayRaw (or marrayNorm) object and then
return the result as second marrayRaw (or marryNorm) object?
Hi Matthew,
passing an marrayRaw into vsn should work as is. Presently the return
value of vsn is essentially an n*d-matrix, with n=number of spots and
d=twice the number of chips, odd-numbered columns containing the green
and
even-numbered columns the red intensities.
There is a certain redundancy between the marrayNorm class and the
exprSet
class, which I suppose the bioconductor programmers want to resolve to
make things easier. In the course of that, the return value of vsn
will
then also be whatever is the merged version of exprSets and
marrayNorm.
For the present, a workaround would be something like the following:
data(swirl)
ns = maNorm(swirl)
vsns = vsn(swirl)
green = seq(1, ncol(vsns@h), by=2)
red = seq(2, ncol(vsns@h), by=2)
maM(ns) = vsns@h[, red] - vsns@h, green]
maA(ns) = (vsns@h[, red] + vsns@h, green])/2
After which ns is an marrayNorm object with slots maM and maA from the
result of vsn. I haven't tested this much, so please make sure it does
want you want. And let me know if it works for you.
Note: if you call it like this, vsn will estimate the normalization
and
variance stabilization parameters simultaneously for the data from all
chips. Alternatively, you could call vsn separately for each chip. I
do
not really have a strong opinion on what is "better".
Best regards
Wolfgang
On Mon, 24 Mar 2003, Matthew Fero wrote:
> Could someone suggest a procedure for performing scale normalization
> with the vsn method on an marrayRaw (or marrayNorm) object and then
> return the result as second marrayRaw (or marryNorm) object?
>
> _______________________________________________
