Identifying nde genes as controls

[ksherwood]

Hi all,

Most of us are interested in identifying differentially expressed genes within datasets, however I was wondering if anyone had tried to identify NON DIFFERENTIALLY EXPRESSED genes- perhaps as a way of picking control genes?

I haven't even got any idea how one would go about doing this? All and any suggestions would be greatly appreciated!

thanks,
k

[Bukowski]

This paper: http://www.biomedcentral.com/1471-2164/12/156

and this R package:

http://www.bioconductor.org/packages/2.2/bioc/html/SLqPCR.html

Which is aimed at identifying the best 'housekeeping' genes for normalisation in qPCR

[ksherwood]

Identifying NDE genes as controls

Thank you very much for that help!

I was wondering if the R package would work for miRNA data? I have a miRNA seq dataset (nanostring) of ~700miRNA and need to identify the least variable (i.e. Non differentially expressed) miRNA's between a disease vs control comparison.

much appreciated!
karen

[steven]

Thank you very much for that help!

I was wondering if the R package would work for miRNA data? I have a miRNA seq dataset (nanostring) of ~700miRNA and need to identify the least variable (i.e. Non differentially expressed) miRNA's between a disease vs control comparison.

much appreciated!
karen

mhm.. if you consider the qPCR world (again) the endogenous control classical approaches (like GeNorm and NormFinder) may not be optimal for miRNAs.
Have a look at this paper:
A novel and universal method for microRNA RT-qPCR data normalization.
The idea is more or less to build a virtual control gene using the median expression. In your case the upper quartile may be more appropriate (like proposed for RNA-seq by Bullard et al I think).

[ksherwood]

Thank you for all the useful pointers!

From a slightly different angle;

Rather than trying to identify a few stably expressed miRNA to use for normalization, are there any suggestions for an approach to identify a NDE profile, thereby giving a unchanged 'background' for your condition of interest? Especially when dealing with NG methods, which can differ from qRT-PCR analytical methods (although eventually findings generally have to be validated in PCR).

It comes from this idea that there should be a 'healthy' profile or baseline, as opposed to a 'disease' profile, which might be more popular as personalized medicine becomes more of a reality.

Thanks again!