add reducedDims to ArchR project from Seurat clustering

[mitchTwoTimes]

Hello,

Is there a way to add Seurat clustering results to the reducedDims slot of an ArchR project? Here is the dimensional reduction analysis I would like to add:

> str(seurat.rna@reductions$global_umap)
Formal class 'DimReduc' [package "SeuratObject"] with 9 slots
  ..@ cell.embeddings           : num [1:65078, 1:2] -8.98 -11.48 -11.62 -8.31 -8.47 ...
  .. ..- attr(*, "scaled:center")= num [1:2] -0.0731 -0.0396
  .. ..- attr(*, "dimnames")=List of 2
  .. .. ..$ : chr [1:65078] "d7_br1#ACCTTCATCTCTAGCC-1" "d7_br1#GGTTGCTCAATGCGCT-1" "d7_br1#CTGGTAAGTAGTTGGC-1" "d7_br1#TGCTTAAAGGAATAAC-1" ...
  .. .. ..$ : chr [1:2] "UMAP_1" "UMAP_2"
  ..@ feature.loadings          : num[0 , 0 ] 
  ..@ feature.loadings.projected: num[0 , 0 ] 
  ..@ assay.used                : chr "originalexp"
  ..@ global                    : logi TRUE
  ..@ stdev                     : num(0) 
  ..@ key                       : chr "UMAP_"
  ..@ jackstraw                 :Formal class 'JackStrawData' [package "SeuratObject"] with 4 slots
  .. .. ..@ empirical.p.values     : num[0 , 0 ] 
  .. .. ..@ fake.reduction.scores  : num[0 , 0 ] 
  .. .. ..@ empirical.p.values.full: num[0 , 0 ] 
  .. .. ..@ overall.p.values       : num[0 , 0 ] 
  ..@ misc                      : list()

Thank you

[rcorces]

There isnt a function to do this but here is some example code that might help:

colData2 <- readRDS("oldMetaData.rds")
all.equal(rownames(colData2), proj2$cellNames)

df <- DataFrame(row.names=proj2$cellNames, "custom#UMAP1" = colData2$UMAP1, "custom#UMAP2" =  colData2$UMAP2, check.names = FALSE)
proj2@embeddings$customUMAP <- SimpleList(df = df, params = list())
proj2@cellColData$Clusters <- colData2$Clusters
plotEmbedding(ArchRProj = proj2, embedding = "customUMAP", colorBy = "cellColData", name = "Clusters")

[mitchTwoTimes]

Thanks for the response. I was able to add my Seurat-generated UMAP coordinates to the embeddings slot and successfully recreate the plot.

However, I actually need to add my Seurat UMAP info to the reducedDims slot to run addPeak2GeneLinks. The reducedDims slot seems a bit different than embeddings - any hints to deal with this?

[rcorces]

ah. my mistake. there are a lot more moving parts with reducedDims. I dont have code to do that and I dont off the top of my head know which components of the reducedDims slot is used in various parts of ArchR. You can try to mimic what its supposed to look like and see if it works.

[mitchTwoTimes]

Hey @rcorces, I wanted to revisit this. I ended up just using the LSI clustering to get a p2g-friendly reducedDims embedding. I colored by my Seurat clusters and they were basically recapitulated in the LSI space.

Would it be catastrophic if ran addPeak2GeneLinks on the LSI space but groupBy Seurat clusters?

[rcorces]

that sounds like it should be fine to me