Tips for finding markers and using outputs of chromVAR motifs

[dgagler]

Hi, thanks for the great package. Seeking assistance for working with differential markers + chromVAR motifs. I've generated a MotifMatrix as per the tutorial and I'm interested in finding differential motifs between 2 sets of clusters. To that end, I ran getMarkerFeatures() with useMatrix = "MotifMatrix" and useSeqnames = "z". I'm struggling to work with the outputs and I'm wondering if I'm missing something. I'll raise 2 specific points.

1. After finding differential markers for other matrix types (like gene scores), I can use getMarkers to extract a set of markers above some threshold and the result is a coherent and accessible dataframe. That does not appear possible with the MotifMatrix. Is that true?

2. Beyond that, interfacing with the output of getMarkerFeatures() with my MotifMatrix is clunky and non-intuitive and I'm not seeing any documentation to assist. I've discovered that I need to go several layers down to access something like MeanDiff values (varDev.markers@assays@data@listData[["MeanDiff"]]) but the result doesn't even provide motif names. Am I missing something?

varDev.markers <- getMarkerFeatures(archR, 
                         useMatrix = "MotifMatrix", 
                         useSeqnames = "z",
                         groupBy = "yesnoPC",
                         bias = c("TSSEnrichment", "log10(nFrags)"),
                         testMethod = "wilcoxon")

> varDev.markers
class: SummarizedExperiment 
dim: 870 2 
metadata(2): MatchInfo Params
assays(6): Mean FDR ... AUC MeanBGD
rownames(870): 1 2 ... 869 870
rowData names(3): seqnames idx name
colnames(2): no yes
colData names(0):

> head(varDev.markers@assays@data@listData[["MeanDiff"]])
            no         yes
1 -0.108513978  0.07016714
2  0.002101452 -0.00675950
3 -0.227610424  0.11373341
4 -0.048518498  0.01461492
5 -0.079875111  0.02459022
6  0.149080310 -0.08015982

I've been poking around the discussion forum and haven't seen any clarification on these points although #1177 and #1279 seem to have found a way around it. Thanks!