Integrating GEX and ATAC peaks : 10x multiome

[shregeno8]

Hi
I am working on a 10X multiome(GEX and ATAC) data and I would really like to use ArchR. Is it possible to integrate the Gene Expression data to the ATAC seq generated from the same 10x multiome experiment?

Thanks!
She

[rcorces]

https://greenleaflab.github.io/ArchR_2020/Ex-Analyze-Multiome.html

[snghprtk]

@ZayaAyush
I am getting the same error as yours with addCommbinedDims. Were you able to solve it?

proj <- addCombinedDims(proj, reducedDims = c("LSI_ATAC", "LSI_RNA"), name = "LSI_Combined")
Filtering 1 dims correlated > 0.75 to log10(depth + 1)

Error in f(init, x[[i]]): number of rows of matrices must match (see arg 2)
Traceback:

1. addCombinedDims(proj, reducedDims = c("LSI_ATAC", "LSI_RNA"),
   . name = "LSI_Combined")
2. lapply(seq_along(reducedDims), function(x) {
   . rD <- getReducedDims(ArchRProj = ArchRProj, reducedDims = reducedDims[x],
   . dimsToUse = dimsToUse, scaleDims = scaleDims, corCutOff = corCutOff)
   . cV <- apply(as.matrix(rD), 2, var)
   . normV <- 1/sqrt(sum(cV))
   . rD * normV * dimWeights[x]
   . }) %>% Reduce("cbind", .)
3. Reduce("cbind", .)
4. f(init, x[[i]])

[doulijun777]

Hello, I also encountered this problem. I successfully do the ATAC and RNA diemension reduction process, and when I do the integrated analysis, it is reported that,

projMulti2 <- addCombinedDims(projMulti2, reducedDims = c("LSI_ATAC", "LSI_RNA"), name = "LSI_Combined")
Error in f(init, x[[i]]) :
number of rows of matrices must match (see arg 2)

Do you have any way to solve this problem?

Thank you so much.

[snghprtk]

@doulijun777 Sorry I haven't figured it out. Waiting for @rcorces and @ZayaAyush to comment.

[ZayaAyush]

sorry, I couldn't make progress with ArchR and switched to the Signac package and analyzed the multiome data well.

[cedmo001]

@snghprtk @doulijun777 Have you figured out a fix for this? I have the same issue!

[ZayaAyush]

@rcorces rcorces
@cedmo001
@snghprtk
Did you figure out this problem? I moved back to ArchR, i have still same issue!

projMulti3 <- addCombinedDims(projMulti3, reducedDims = c("LSI_ATAC", "LSI_RNA"), name = "LSI_Combined")
Error in f(init, x[[i]]) :
number of rows of matrices must match (see arg 2)

Here is my scripts:
inputFiles <- getInputFiles("~/Documents/ArchRanalysisofscATAC")
inputFiles <- c("~Documents/ArchRanalysisofscATAC/WT_atac_fragments.tsv.gz",
"~Documents/ArchRanalysisofscATAC/cKO_atac_fragments.tsv.gz")
names(inputFiles) <- c("WT", "cKO")

ArrowFiles <- createArrowFiles(inputFiles, force = TRUE)
ArrowFiles <- c("WT.arrow","cKO.arrow")

projMulti2 <- ArchRProject(ArrowFiles = ArrowFiles, outputDirectory = "~/Documents/ArchRanalysisofscATAC")

#Import scRNA

input <- c("/Documents/file1/filtered_feature_bc_matrix.h5",
"/Documents/file2/filtered_feature_bc_matrix.h5")

#set your names variable
names <- c("WT", "cKO")

#set your featureType variable
featureType <- "Gene Expression"

featureMats <- lapply(seq_along(input), function(y) {
message("Importing Feature Matrix ", y, " of ", length(input))
ArchR:::.importFM(featureMatrix = input[y], featureType = featureType,
name = names[y])
})

seRNAcombined<-cbind(assay(featureMats[[1]]), assay(featureMats[[2]]))

seRNA<-SummarizedExperiment(assays=list(counts=seRNAcombined), rowRanges= rowRanges(featureMats[[1]]))

projMulti2 <- addGeneExpressionMatrix(input = projMulti2, seRNA = seRNA)

projMulti2 <- addDoubletScores(projMulti2, force = TRUE)
projMulti3 <- filterDoublets(projMulti2)
	
## Analysis of multiome data in ArchR

projMulti3 <- addIterativeLSI(
  ArchRProj = projMulti3, 
  clusterParams = list(
    resolution = 0.6, 
    sampleCells = 10000,
    n.start = 10
  ),
  saveIterations = FALSE,
  useMatrix = "TileMatrix", 
  depthCol = "nFrags",
  name = "LSI_ATAC",force = TRUE
)

projMulti3 <- addIterativeLSI(
  ArchRProj = projMulti3, 
  clusterParams = list(
    resolution = 0.6, 
    sampleCells = 10000,
    n.start = 10
  ),
  saveIterations = FALSE,
  useMatrix = "GeneExpressionMatrix", 
  depthCol = "Gex_nUMI",
  varFeatures = 2500,
  firstSelection = "variable",
  binarize = FALSE,
  name = "LSI_RNA",force = TRUE
)
projMulti3 <- addCombinedDims(projMulti3, reducedDims = c("LSI_ATAC", "LSI_RNA"), name =  "LSI_Combined")
Error in f(init, x[[i]]) : 
  number of rows of matrices must match (see arg 2)

[dannyconrad]

@ZayaAyush I just ran into this problem as well and I think I have the solution - at least for my case. As the error message suggests, the problem was that the dimensions of the two SVD matrices you're trying to combine are not identical. You can check this like so:
projMulti3@reducedDims$LSI_ATAC$matSVD %>% dim()
projMulti3@reducedDims$LSI_RNA$matSVD %>% dim()

The number of rows should in theory be the same for both and should match the number of cells in the ArchR project. I think in my case, the problem was that there were cells with 0 or otherwise very few RNA counts - and these were getting removed at some point in the IterativeLSI process - likely because they would otherwise break the computation.

So try first filtering your project to remove cells with <5 or <10 RNA counts, these won't be very informative to your analysis anyway.

[ZayaAyush]

Thank you @dannyconrad
I checked the SVD matrics, and they are different.
projMulti3@reducedDims$LSI_ATAC$matSVD %>% dim()
[1] 14749 30
projMulti3@reducedDims$LSI_RNA$matSVD %>% dim()
[1] 14733 30

Can you please provide instructions on how to obtain 14733 cells from both data sets to be filtered out?
Thank you for your advice,