JThomasWatson
commented
4 hours ago @gfinak @amcdavid Have you seen this behavior before?
Open JThomasWatson opened 2 weeks ago
JThomasWatson
commented
4 hours ago @gfinak @amcdavid Have you seen this behavior before?
If you have a question about how to use MAST, or how to interpret the output please submit a question to the bioconductor support site so that others can benefit from the discussion. Please do not use github issues.
Otherwise Please briefly describe your problem and what output you expect.
Please include a minimal reproducible example (AKA a reprex). If you've never heard of a reprex before, start by reading https://www.tidyverse.org/help/#reprex.
Brief description of the problem
I'd like to use MAST with sample ID as a random effect. When I include anything in the
hookparameter as in section 4.2 of the MAIT tutorial, the ZlmFit object becomes full ofNAs and no residuals are accessible. In the summary table created bymsummary <- summary(zlmResidDE, doLRT = lrt_name, parallel = TRUE), all fold change values areNaNwhile everything else is NA. Below is the code I use to attempt a zlm model with residual hook.And the produced ZlmFit object looks like this:
Where most values are
NA, and thehookOutslot isNULLfor each gene.When I don't include
hook, the ZlmFit object is populated with data, and the summary table builds properly, but no residuals are stored. The code used to build the model without residual hook is:Viewing the ZlmFit object created by that code shows this:
I attempted to run the residual hook function line by line to better understand what's happening, and I noticed warnings after the commands
class(x@fitC) <- c("glm","lm")andclass(x@fitD) <- c("bayesglm","glm","lm")that said these objects would no longer be S4 objects. I wondered if that class change was breaking a subsequent operation that expected an S4 object, so I attempted to run a modified version by cloning this repository, building and loading the package locally, and redefiningcombined_residuals_hook()without those two class change commands, but that doesn't appear to have solved it, as including thehookparameter still introducesNAs and doesn't store residuals. Below is my output fromsessionInfo():