Hi,

in the fourCseq package, in the getDifferences function, if fitNormFactors is set to FALSE, then the dispersion estimation and the binomialWaldTest are performed on raw reads counts? Or on a sequencing-depth normalized count like usually in DESeq?

When looking for difference between samples, as we do not expect any influence of the distance to the view point for far away and trans interactions, is it possible to just call differential counts with DESeq2? Do you see any reason not to do so? 

Hello Samuel,

if fitNormFactors is set to FALSE the dispersion estimation and the binomialWaldTest are performed on the normalized read counts, which are normalized for sequencing depth like usually in DESeq2.

You can try to look for differences in samples far away from the viewpoint using DESeq2 directly, but I would expect that the data is quite noisy in this region. So you definitely have to be careful with interpreting your results.

Cheers, Felix

Samuel

I don't think we have enough experience to guide you regarding your question on very distal or trans interactions. It would be interesting to hear what you find. From my understanding of the biology, highly specific point-to-point interactions between the viewpoint and one single locus on a fragment very far away will be rather rare, and more frequent will be somewhat more diffuse 'scanning'(?) interactions. Any analysis should probably anticipate that.

Wolfgang

Hello Samuel,

trafo is the assay with the values of the variance stabilizing transformation.

You can get the normalized count data (normalized with all provided factors up to this point) with counts(fc, normalized=TRUE).

Felix