[BioC] Error in ebayes: No residual degrees of freedom in linear model fits

[Lucas Santana dos Santos]

Dear Wolfgang,

I will try to use BGX in order to identify differentially expressed  
genes. Seems to me it can detect differentially expressed genes  
without replicates (it is worse than Limma with replicates)..
Also I will follow your suggestion to ask a more specific question  
about my data. My main interest is to find classifiers, but I guess I  
need to know which genes are differentially expressed to do so.

Thank you.

Best,

Lucas


On May 26, 2009, at 5:38 PM, Wolfgang Huber wrote:

> Lucas Santana dos Santos ha scritto:
>> Dear Group,
>> I am trying to find differential expressed genes using Limma,  
>> however I got the following error when I try eBayes:
>> Error in ebayes(fit = fit, proportion = proportion, stdev.coef.lim  
>> = stdev.coef.lim) :
>> No residual degrees of freedom in linear model fits
>> My dataset consists of 8 arrays from 8 different cell lines.  I  
>> have no replicates.
>> I just want to see which genes are differentially expressed between  
>> these cell lines.
>> I am wondering if I need more than one array per cell line in order  
>> to make this work.
>> Maybe eBayes need replicates to generate variability. Does it make  
>> sense?
>> A similar post already exists: https://stat.ethz.ch/pipermail/bioconductor/2005-May/009027.html
>> Thank you,
>> Lucas
>
> Dear Lucas
>
> without replication, you will not be able to use limma / eBayes to  
> get statistical measures of significance of differential expression.
>
> What you can do is:
> - look at the width of distribution of (log-)intensities and eye- 
> ball the large-looking ones, or
> - decide on a more specific question to ask from your data, or
> - ask the people who produced the data to provide replication.
>
> Best wishes
>     Wolfgang
>
> ------------------------------------------------
> Wolfgang Huber, EMBL, http://www.ebi.ac.uk/huber