generating bedgraphs using RUVseq normalisation

[ahorvath]

Dear @drisso,

I would like to generate genome-wide (e.g. on 10b-window) bedgraphs using the normalisation factor/linear equation calculated/fitted by RUVSeq.
Do I have to use W1, W2, etc and the offset parameter?
Is that a library size normalisation also neccessary?

Many thanks.
Bests,
Attila

[drisso]

Dear Attila,

I'm not sure if I completely understand your question. What do you want to visualize in a bedgraph exactly? The normalized values after the correction? If so, the easiest way would be to extract the normalizedCounts values, as specified in the vignette.

Also in the vignette, you will see how, for gene expression analysis, we recommend library size normalization prior to removing unwanted variation with RUV. However, it is not clear to me if that is beneficial in your genome-wide analysis: RUVSeq was not designed with such an application in mind, so you should carefully look at the results to see which option is preferable in your case.

Best,
Davide

[drisso]

Also, please consider expanding on your question by providing more background on what you are trying to achieve with your analysis and consider posting this type of questions on support.bioconductor.org

[ahorvath]

Many thanks for your answer.

The reason why I would like to normalise for spikes-in is that I am interested in non-coding RNAs genome-wide (including eRNAs that are typically on intergenic regions). I thought that it would be a good idea to normalise for the CDS or the whole mRNA regions using RUVSeq and apply the scale factor for the entire genome. Also, when showing snapshots of individual examples, it would be good to see the same normalised values on the bedgraphs/bigwigs as the ones in the tables we work with.

By library normalisation, do you mean the betweenLaneNormalisation step prior to RUVseq?

Bests,
Attila