search

mervesa / HiCDCPlus

HiCDCPlus
15 stars 2 forks source link

fitType parameter for hicdcdiff() #5

Closed jessakay closed 2 years ago

jessakay commented 3 years ago

Thanks for making this package available along with some very helpful documentation. I'm having trouble understanding the parameter choice for fitType in DESeq2::estimateDispersions() that is called by hicdcdiff(). The default here seems to be local, different from the original default value of parametric in DESeq2 itself, and the tutorial here seems to use mean instead. Could you please share the rationale behind the choices made here and when one may be more appropriate than the others?

mervesa commented 2 years ago

Hi @jessakay , I would refer to the Bioconductor discussion here by the package's author. I've found local generally forms a more refining fit suitable to the overdispersed nature of interaction counts data, hence that's the default. This may change by resolution and read depth. If you want to check which choice is appropriate, it is a good idea to get the mean-dispersion plots by choosing diagnostics=TRUE, and following the guidelines in this and similar threads (e.g., https://support.bioconductor.org/p/81094/).

mervesa commented 2 years ago

I'm closing this because of inactivity.