[BioC] EBarrays question

[jgomez at uni-potsdam.de]

Dear List,
I will describe my problem in more detail. Thanks Naomi for your  
comments. Below you'll fin the R script, output and info about my  
experiment. I hope someone help me to figure out the problems.
I have two sampling time points, morning and afternoon, and twelve  
different samples (let's say different tissues), so I have 2*12=24  
MAs. The samples are neither biological nor technical replicates, they  
are as I wrote already different samples.
My object is a data matrix of the kind "ExpressionSet", meaning I have  
normalized intensity values (of course no negative values in the data  
matrix).

R script:
> library(EBarrays)
> pattern<-ebPatterns(c("1,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0","1,0,0,0,0,0,0,0,0,0,0,0,2,0,0,0,0,0,0,0,0,0,0,0"))
#With this pattern I hope to find DE genes between AM and PM time  
points in #condition sample one. As I understand "1,1" is the NULL  
expression pattern, #while "1,2" counts for µ AM different from µ PM,  
sample 1.

> pattern
Collection of 2 patterns

     Pattern 1 has 1 group
        Group  1: 1 13

     Pattern 2 has 2 groups
        Group  1: 1
        Group  2: 13
> gg_data.out<-emfit(data,family="GG",hypotheses=pattern, num.iter=20)
> gg_data.out

  EB model fit
          Family: GG ( Gamma-Gamma )

  Model parameter estimates:

              alpha   alpha0         nu
Cluster 1 357.0943 4.742546 0.05231341

  Estimated mixing proportions:

           Pattern.1  Pattern.2
Cluster 1 0.9838123 0.01618772

> gg.post.out<-postprob(gg_data.out, data)$pattern
> gg.post.out
NULL

I get the same result for other patterns or using not GG but LNN.
To test if a less complex dataset would work better, I used another  
time course. The experiment is a time course of 6 MAs, two genotypes  
(mutant and wild type) sampled at 0, 12 and 48 h after treatment. The  
results are identical, NULL for posterior probability.
I have tried to contact M. Yuan and C. Kendziorski, but got no answers  
to my mails.
If anyone can point me to the mistake or mistakes I am doing, or can  
suggest another program available for R or standalone to analyse time  
series without replicates, I will be very, VERY grateful.
Regards,
Judy


-- 
Judith Lucia Gomez, PhD
Universität Potsdam
BioPhysics and Molecular Plant Biology
Heisenberg-Group
D-14476 Potsdam
Germany


Quoting Naomi Altman <naomi at stat.psu.edu>:

> p.s. You need to post your code and sessionInfo to get help.
>
>
> At 05:08 AM 4/8/2011, you wrote:
>> Dear List,
>> I would like to use EBarrays to analyse 24 MAs without replicates. I
>> have 12 samples taken at one time point, and 12 taken at other time
>> point. I want to find the genes differentially express at one time
>> point vs. the other.
>> I have chosen EBarrays because I don't have replicates.  Until the
>> fitting of the model (emfit) it seems to work. However, when I want to
>> check the post. probabilities I get NULL. I have tried all kinds of
>> things I could think of. The answer is always:
>>
>> gg_post_IA_P01.out
>> NULL
>>
>> if I understand right, most genes belong to the µ1=µ2 pattern
>> Pattern.1  Pattern.2
>> Cluster 1 0.9838123 0.01618766
>>
>> But no matter how I set the pattern (including all MAs, comparisons of
>> only one time point at a time, etc), the answer is always NULL for
>> post. prob.
>> Does anyone has any idea how to set my analysis, or at the very least
>> why the calculation of post. probs always renders NULL?.
>> Thanks in advance.
>> Regards,
>> Judith
>>
>>
>>
>> --
>> Judith Lucia Gomez, PhD
>> Universität Potsdam
>> BioPhysics and Molecular Plant Biology
>> Heisenberg-Group
>> D-14476 Potsdam
>> Germany
>>
>> _______________________________________________
>> Bioconductor mailing list
>> Bioconductor at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/bioconductor
>> Search the archives:   
>> http://news.gmane.org/gmane.science.biology.informatics.conductor