[BioC] EBarrays question

[Hervé Pagès]

Hi Judy,

I'm not sure the EBarrays authors/maintainers are following the
Bioconductor mailing lists. You might want to try to contact them
directly for this.

Cheers,
H.


On 11-04-11 03:39 AM, jgomez at uni-potsdam.de wrote:
> 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
>
>


-- 
Hervé Pagès

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M2-B876
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319