Compatibility with R 3

[PedroMilanezAlmeida]

Just wanted to check if any version of this amazing package was ever compatible/installable with R 3.6. Any help would be much appreciated.

EDITED: in case the above is true (it seems to be according to original pub [R 3.6.2, glmGamPoi 1.1.5]) and the package is usable/installable in R 3.6, which release/commitSHA should be used with devtools::install_github, remotes::install_github or some other installation tool?

[const-ae]

Hey Pedro,

Thank you for your nice comment.

The only feature from 4.0+ I am using is the anonymous function shorthand notation (lapply(x, \(e) ...). I just checked I only use it in the pseudobulk function. So if you go with the last stable release (Bioc 3.16, that is commit 048e17384209fc07031e09875ec6eea35e90ef46), I think you should be fine.

The more challenging problem might be the dependencies that are also on Bioconductor (DelayedMatrixStats, SummarizedExperiment etc.). I assume you might be able to install them as well using devtools but I cannot tell you if the latest GitHub version will work or if you will also have to go back a few versions.

Best, Constantin

[PedroMilanezAlmeida]

Hey Constantin,

Thanks for the detailed answer! Your package is just that... amazing!

We decided to go with version 1.1.7 (remotes::install_github('const-ae/glmGamPoi@12449cd')) since, unfortunately, the code is not tagged and that built via github actions (and is close to the paper version of 1.1.5).

At the core of what we need is to return the randomized quantile residuals after fitting the models (EDIT: for single cells). Do you think you would be able to roughly say if updates since v1.1.7 will lead significant differences in the rqr?

Thanks again and best, Pedro

[const-ae]

Version 1.1.7 is pretty old. I don't know if anything significant has changed in the last 2.5 years. A quick look at the blame for the qres.gampoi function suggests that nothing has changed (https://github.com/const-ae/glmGamPoi/blame/master/R/residuals.R#L77)

since, unfortunately, the code is not tagged and that built via github actions

Yeah, unfortunately, the tags added by Bioconductor are somehow not displayed by Github. But you could check out the Bioconductor repository (e.g., for release 3.13) to quickly navigate the different versions.

---

In the end, if you are stuck on such an old version of R there is only so much I can do to help. (Happy to hear though that you found a version that works :) )

[PedroMilanezAlmeida]

I know, right... But that's the reality at the moment. I appreciate your quick and detailed response!