[BioC] SAM output in siggenes package

[James MacDonald]

First, you have (if I did my math correctly) 2^16 combinations that you
could do, but you only did 200. This will make the null distribution
pretty granular, so you might want to consider increasing B to something
like 1000 or so. Realistically you *should* do more than 1000, but you
will run into time issues pretty quickly (oddly enough siggenes was much
faster in R-1.6.1 and appeared to slow down appreciably around R-1.7.1
or so. Why this is, I don't know).

Note that most of the 'significant' genes you are picking up may not
actually be significant in the biological sense. Then again, maybe they
are. One problem (as I see it) with microarray data analysis is that we
are making a strong assumption that gene expression is more or less
directly related to phenotype. However, I think some genes probably get
a lot done with very little change in expression whereas other genes may
require a huge change in expression to effect a moderate change in
phenotype.

All this aside, many people require an additional fold change constraint
for the genes they list as being significant. This will make most if not
all of your 'significant' genes go away.

As for the warning messages, these indicate that you are using an old
version of siggenes that is not quite compatible with the current
version of R. I would highly recommend that you upgrade to the current
release version of BioC.

HTH,

Jim



James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623
>>> "Anna Cao" <yunie at caltech.edu> 12/22/03 8:02 PM >>>

Hi,

I recently perform SAM using the siggenes package. I'm doing SAM on
untreated vs untreated cultured for 1 hour sample. I have 16 replicates
and
am doing the paired two-class SAM. Ideally, there should be no
significant
genes but SAM in R gave me a lot of significant genes (~6000 genes with
~0%
FDR). I counted the number of 2 fold genes and they are less than 100.
Are
there any parameters that I missed and need to adjust b/c this result
seems
incorrect.

Here's my input and output:
> sam.UT60<-sam(RG.UT60,
x=c(1:16),y=c(17:32),paired=TRUE,B=200,na.rm=TRUE)
Warning: There are 7958 genes with at least one missing value.
 
s0 = 0 
 
SAM Analysis for a set of delta: 
   delta    p0   false called   FDR
1    0.2 0.245 10801.0  13948 0.189
2    0.4 0.245  8877.0  13385 0.162
3    0.6 0.245  6079.5  12508 0.119
4    0.8 0.245  3797.0  11577 0.080
5    1.0 0.245  1704.0   9837 0.042
6    1.2 0.245   812.0   8630 0.023
7    1.4 0.245   391.0   7671 0.012
8    1.6 0.245   182.5   6846 0.007
9    1.8 0.245    85.0   6091 0.003
10   2.0 0.245    39.0   5312 0.002

Warning messages: 
1: multi-argument returns are deprecated in: return(r, s, r.perm,
s.perm, Z,
mat.samp, var.0.genes, NA.genes) 
2: multi-argument returns are deprecated in: return(alpha.hat, s.zero,
cv,
cv.zero) 
3: multi-argument returns are deprecated in: return(p0, spline.out,
vec.p0) 
4: multi-argument returns are deprecated in: return(tab.fdr, mat.fdr,
p0) 
5: multi-argument returns are deprecated in: return(d, d.sort, s, d.bar,
d.perm, mat.samp, s0, FDR, p0, fdr.ngenes,  

BTW, why am I getting these warning messages?

THANKS! 

Anna

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