Can I use custom feature selection for addIterativeLSI?

[TobyHoldcroft]

Thanks for creating such a powerful tool!

I am analysing a 10X multiome dataset and have done some preliminary work using scverse tools to do QC and deviance-based feature selection on my scRNAseq component. I have used ArchR for scATACseq QC and intend to use it later for peak calling and further analysis.

I want to use addIterativeLSI to visualise my RNA, ATAC and combined dims, however I don't know how to choose a custom vector of my selected RNA features from deviance modelling. Is this currently possible using addIterativeLSI?

As a workaround I was thinking I could subset my GeneExpressionMatrix to only those genes in my selected features list, but then I can only use up to half of those due to the requirements of varFeatures. Perhaps I can add 'pseudo-features' with 0 counts so that my top features come out as my selected features? But this method seemed a bit contrived and I assume other people might also want to use custom selected features in a more streamlined way if there is one.

Any guidance is appreciated.
Many thanks