[BioC] Newbie question regarding SAM analysis

Simon Lin simonlin at duke.edu
Sat Dec 24 19:17:59 CET 2005


Hi Nick:

You are right. There is something weird about your data. Did you 
generate a simple pairwise plot (MvA or a simply scatter plot) of any 
pair of the arrays?

 From the SAM output, it seems that there is no differentially expressed 
genes. FDR are all close to 1, no matter what delta is, which means 100 
percent identified (called) genes are false positive.

There is only 18 called genes, even though delta is already very small, 
and the falsely identified one become larger than identified genes!

Simon

Message: 3
Date: Thu, 22 Dec 2005 11:51:55 -0600
From: "Ettinger, Nicholas" <nicholas-ettinger at uiowa.edu>
Subject: [BioC] Newbie question regarding SAM analysis
To: <bioconductor at stat.math.ethz.ch>
Message-ID:
	<A4AA05CE92DACC43886D133DB94A926E218CD063 at medicine-exch1.medicine.uiowa.edu>
	
Content-Type: text/plain

Hello all!

 

This is my first post.  Any help or suggestions would be greatly
appreciated!

 

I am trying to analyze 8 arrays (4 untreated, 4 treated; paired &
alternating) with SAM.

When I read the vignettes from 'siggenes' and looked at the sample
diagrams, I was expecting to see my 'Called' column go from some number
much closer to the number of probes on the hgu133probe2 Affy gene chip
(something like 50,000 I think) down to zero.  Why does it only start at
18?

 

I am thoroughly confused by that.

 

Thanks for any suggestions!!

 

Happy Holidays to all!!

 

---Nick Ettinger

University of Iowa

 

Here is my code:

TotalData <- ReadAffy()

chipnumber <- length(sampleNames(TotalData))

chipnames <- sampleNames(TotalData)

eset_rma <- rma(TotalData)

 

K <- chipnumber/2

eset.cl <- rep(1:K, e = 2) * rep(c(-1, 1), K)

eset.gnames <- geneNames(TotalData)

sam.out <- sam(eset_rma, eset.cl, rand = 123, gene.names = eset.gnames)

sam.out

 

SAM Analysis for the Two-Class Paired Case  

      Delta p0    False       Called      FDR

1     0.1   0.986 44.500      18          1

2     0.3   0.986 28.250      13          1

3     0.4   0.986 3.688       2           1

4     0.6   0.986 1.062       1           1

5     0.8   0.986 1.062       1           1

6     1.0   0.986 1.062       1           1

7     1.1   0.986 1.062       1           1

8     1.3   0.986 1.062       1           1

9     1.5   0.986 1.062       1           1

10    1.6   0.986 1.062       1           1

 

summary(sam.out)

 

SAM Analysis for the Two-Class Paired Case  

 s0 = 0.0646  (The 10 % quantile of the s values.) 

 Number of permutations: 16 (complete permutation) 

 MEAN number of falsely called genes is computed.

 

      Delta       p0          False       Called FDR  cutlow      cutup
j2    j1

1     0.1        0.986       44.500      18    1     -4.503      6.912
16    54674

2     0.3        0.986       28.250      13    1     -5.004      6.912
11    54674

3     0.4        0.986       3.688       2     1     -7.102      Inf
2     54676

4     0.6       0.986       1.062       1     1     -8.690      Inf
1     54676

5     0.8       0.986       1.062       1     1     -8.690      Inf
1     54676

6     1.0       0.986       1.062       1     1     -8.690      Inf
1     54676

7     1.1       0.986       1.062       1     1     -8.690      Inf
1     54676

8     1.3       0.986       1.062       1     1     -8.690      Inf
1     54676

9     1.5       0.986       1.062       1     1     -8.690      Inf
1     54676

10    1.6         0.986       1.062       1     1     -8.690      Inf
1 5   4676



More information about the Bioconductor mailing list