[BioC] analysis with no replication

Richard Pearson richard.pearson at postgrad.manchester.ac.uk
Tue Jun 12 12:48:24 CEST 2007


I'm not sure if I've understand you fully but if you are wanting to 
identify differentially expressed genes from unreplicated GeneChip data 
you have a couple of options that I am aware of:

1. The puma package can be used to determine uncertainties in expression 
levels which can then be used to determine probabilities of differential 
expression in a principled way. I've not included an unreplicated 
example in the puma vignette, but this is quite straightforward to do so 
give me a shout if you'd like to do this with puma but can't figure out 
how. I believe the bgx package can also do something similar.
2. You could simply apply your favourite summary method, and then use 
fold change between conditions to determine differentially expressed genes.

Having said the above, with unreplicated data you have no information on 
biological variability, so the ideal approach would be to repeat the 
experiment if at all possible.

Hope that helps


Daofeng Li wrote:
> Hi all,
> i am doing a Medicago GeneChip with six treated time points without
> any replication
> i used limma to analysis that,but the statistic part of Bayesian is
> not avialable in my condition.
> and i wanna got the expression curve of each probeset by the time
> point and select my interested gene with always up regulated for
> example.Could anyone give some advice on how to do analysis on my
> condition? for i am really confused.
> Thank you

More information about the Bioconductor mailing list