Michael Love

@dn070017 Can you provide `sessionInfo()`? Also, my first check would be to just look at the aligned reads with IGV. You can color by strand.

The argument 'strand_specific' is in the devel branch (Polyester 1.9), this is release. Check the help pages in R, and you should see that 'stand_specific' is not listed in the...

I think the broader consequences should certainly be taken into account, and maybe “fixing” this behavior reported above is not worth the downstream effort of many package maintainers.

Counts are modeled as integers in the C++ fitDisp() function, but that's the key place that I see in the codebase. @const-ae would glmGamPoi help or do you also model...

Ok, so I could potentially allow override of the int conversion and pass doubles to glmGamPoi for all estimation. That's cleaner than the alternative

Just a heads up, I plan to work on this side case, e.g. override integer checks so that one can use glmGamPoi within DESeq(), but it may not happen before...

Not yet. But you can directly use glmGamPoi bypassing DESeq2 for the mean time. Or edgeR or limma. I have to see if the proposal is possible without doing a...

You should now be able to avoid integer conversion using `DESeqDataSet()` with `skipIntegerMode=TRUE`, followed by `fitType="glmGamPoi"` ``` > mat se dds dds

@itamburi or @frederikziebell could you give devel branch a try?

hi Christian, Mostly the glmGamPoi is desired for speed, and glmGamPoi requires LRT. I think the default would be fine as well. What I've mentioned on the support site --...