[BioC] DESeq / DESeq2 for translation efficiency in polysome profiling or ribsome profiling experiments

Simon Anders anders at embl.de
Tue Dec 24 15:08:14 CET 2013

Dear Gilgi

On 24/12/13 09:50, Gilgi Friedlander wrote:
> 1.Would you suggest to do the correction of the library size
> (estimateSizeFactors) for all samples together (although there are 2
> types: all mRNAs and ribosome bound mRNAs, that are only a fraction of
> all mRNAs in the cell)? If not, how would you suggest to do the
> normalization?

Not necessary. If you do the normalization separately, the average ratio 
of total-RNA to the bound-RNA libraries will be in the size factors; if 
you normalize all together, it will be absorbed in the "assay" factor. 
In any case, your final result will be the same, as this extra factor 
cancels out in the double ratio you mention below.

> 2.Would you suggest to calculate the ratio: (wt-ribosome/wt-total) /
> (mutant-ribosome/mutant-total) and look at it together with the p-value
> of the interaction assay:treatment?

For the interaction, you get not only a p value but also a log2 fold 
change in the results data frame. The latter is an estimate of the log2 
of the double ratio you mention. Using this estimate is preferrable to 
simply using the ratio itself because it has undergone empirical-Bayes 
shrinkage, which gets rid of the exaggerated values that the naive ratio 
tends to exhibit when counts are low.


More information about the Bioconductor mailing list