Question: EBarrays question
0
8.7 years ago by
jgomez@uni-potsdam.de30 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 ?1=?2 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 Universit?t Potsdam BioPhysics and Molecular Plant Biology Heisenberg-Group D-14476 Potsdam Germany
ebarrays • 543 views
modified 8.7 years ago • written 8.7 years ago by jgomez@uni-potsdam.de30
0
8.7 years ago by
jgomez@uni-potsdam.de30 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 ? AM different from ? 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 Universit?t 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 ?1=?2 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 >> Universit?t 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 ADD COMMENTlink written 8.7 years ago by jgomez@uni-potsdam.de30 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 ? AM different from ? 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