Motif enrichment in a set of peaks

[AnjaliC4]

Hi @rcorces,

I know we can do motif enrichment in markerPeaks object obtained from ArchR which is a SummarizedExperiment object. If we have done a differential analysis of marker peaks elsewhere - we usually get a data.frame with a set of peaks and foldchange and FDR values. How do I convert this into a format similar to the SummarizedExperiment object accepted by peakAnnoEnrichment() for performing motif enrichment? Please see that the peaks were called with ArchR so are present in the peak matrix.

Thanks

[rcorces]

ArchR doesnt have a function to do this (maybe it should). But this can be accomplished with other packages like motifmatchR

[Zepeng-Mu]

I happen to be doing something similar, and I construct my own sce from the peaks I have and use this in peakAnnoEnrichment, I think this should work:

mySe <- SummarizedExperiment(
  assays = list(z = de_res$z, FDR = de_res$FDR, logFC = de_res$logFC),
  rowData = as.data.frame(peakSet),
  metadata = list(Params = list(useMatrix = "PeakMatrix"))
)

enrichMotifs <- peakAnnoEnrichment(
  seMarker = mySe,
  ArchRProj = projPBMC,
  peakAnnotation = "CisBP2021",
  cutOff = "FDR < 0.1"
)

[AnjaliC4]

Hi @Zepeng-Mu, Does the above strategy work for you? I think seMarker has to have all the peaks present in project peakSet for this to work?

[Zepeng-Mu]

This works for me. For the example above, I also have all the peaks. I think this is mainly because when you construct an sce object, you have to include matrices in assays, and obviously matrices needs to have the same shape, hence all the peaks, and then you filter peaks in each condition (column) of the sce by cutOff in peakAnnoEnrichment.

I also have cases where the candidate peaks are just a subset of all peaks in the peakSet, in this case if you just have one condition I think you can just use a one-column matrix in assays, although it's not all the peaks?
In my case, I constructed a new matrix (peakBinary for example) where rows are all peaks and columns are conditions, and then I use 1/0 in the matrix to indicate if the peaks belong to my candidate peak set in each condition. And then I would use something like peakBinary == 1 in cutOff to test for enrichment on those candidate peaks.

Does this answer your question?