Open nfancy opened 7 months ago
Hi,
Thanks the detailed report. What model are you fitting in model_formula
?
Apologies, for some reason, I didn't get a notification of your response. The model formula was ~cngenseon+(1|manifest) + diagnosis
Here's a response I got from Matrix developer.
https://r-forge.r-project.org/tracker/?func=detail&atid=294&aid=6835&group_id=61
However, it did not solve the issue. Thanks.
@nfancy I think this is a binary compatibility issue that might be resolved by recompiling (and/or reinstalling?) some packages. Using the data you provided and a bioconductor 3.17 docker image with installed MAST I could run your model successfully. I'm sorry that I can't provide more guidance than "it works on my machine". The sessionInfo of this container is listed below -- It does have MAST 1.28.0 (which shouldn't matter), but otherwise the same versions of Matrix and lme4 as you reported above. I do see that you are on an ubuntu 18.xx versus ubuntu 20.xx in my container, which could cause some subtle issues with compiled code, I suppose.
Hi, Thanks for checking. Can you think of any dependencies for MAST which depends on Matrix?
lme4 depends on matrix, and is being used in the example you provided. Did you reinstall lme4 from source? I sometimes find these issues require deleting packages manually from the R package library and reinstalling, which has its own pitfalls. Otherwise, you could work in one of the docker containers I linked to above (kind of annoying I realize).
Thanks. We have incorporated MAST in our own package scFlow, hence it's mandatory that we sort this out. :(
You are right, the culprit is lme4 since with method = "glm" it runs file. Unfortunately deleting and reinstalling lme4 with type = "source" did not help.
Oh nice! Since it's nextflow-based, I suppose you could just run the process that references MAST in a docker image, though that would probably constrain your users too much. What environments are you getting the failure? Just on local workstations, or in CI?
We have both an R package with MAST-wrapper and nextflow-based pipeline, hence, we first need to find a fix within our system. The worst case scenario, we may need to wait until the next version release of lme4.
Hi,
Thanks for this package, I am getting this error with the MAST::zlm function. I am scratching my head and can't find a solution. Any pointer is hugely appreciated.
The file was a little to big to attach here so I uploaded the reproducible data here.
https://www.dropbox.com/scl/fi/2zoiot4zx0ntyt4cg1ff0/mast_sca.zip?rlkey=xjkb7588ip38rq7khwnlge60u&dl=0