getMarkerFeatures() error: error in evaluating the argument 'values' in selecting a method for function 'Rle': Object 'PeakMatrix/chrX/jValues' does not exist in this HDF5 file.

[zrcjessica]

Hello,

I have snATAC-seq from multiple samples, and have already labeled the cell types and called peaks on the dataset. Within each cell type, I want to find differential marker peaks between groups A and B (high addiction vs. low addiction).

My command is as follows:

# subset for cells belonging to specific cell type
idxSample <- BiocGenerics::which(proj2$Clusters2 %in% names(table(proj2$Clusters2))[1])
cellsSample <- proj2$cellNames[idxSample]
proj2[cellsSample, ]

markersPeaks <- getMarkerFeatures(
    ArchRProj = proj2[cellsSample, ], 
    useMatrix = "PeakMatrix", 
    groupBy = "Addiction",
    useGroups = "high",
    bgdGroups = "low",
  bias = c("TSSEnrichment", "log10(nFrags)", "Library", "Sample"),
  testMethod = "wilcoxon"
)

On an earlier run, getMarkerFeatures() finished successfully, but the job crashed because I was trying to save the output to an rds file in a directory I had not yet created... oops. I resubmitted the job, but this time it crashed with the following error:

Error in h(simpleError(msg, call)) : 
  error in evaluating the argument 'values' in selecting a method for function 'Rle': Object 'PeakMatrix/chrX/jValues' does not exist in this HDF5 file.
Calls: getMarkerFeatures ... Reduce -> .safelapply -> lapply -> FUN -> Rle -> h5read

This was odd, since on the previous run it didn't have this issue. In fact, I took a look at the ArchR Log file for the previous job where getMarkerFeatures() completed without error, and it contained data for chrX. I'll attach it here, but I'll also paste the last bit of that log here:

2022-01-12 20:22:19 : high_chrX_diffResult, Class = data.frame
high_chrX_diffResult: nRows = 5965, nCols = 8
        log2Mean      log2FC       fdr        pval mean1 mean2   n      auc
f472137 3.169925 -0.65207670 0.7458748 0.257314532 0.012 0.020 500 0.497010
f472138 3.321928 -0.58496250 1.0000000 0.757716124 0.014 0.022 500 0.498980
f472139 4.087463  0.16992500 1.0000000 0.639562933 0.034 0.030 500 0.501972
f472140 4.209453 -0.07800251 0.9684668 0.495822492 0.034 0.036 500 0.497056
f472141 5.169925 -1.37851162 0.7458748 0.007900311 0.038 0.102 500 0.484984
f472142 1.000000  1.58496250 0.7458748 0.157506917 0.004 0.000 500 0.502000


###########
2022-01-12 20:22:21 : Completed Pairwise Tests, 0.83 mins elapsed.
###########

------- Completed

End Time : 2022-01-12 20:22:22
Elapsed Time Minutes = 0.870024279753367
Elapsed Time Hours = 0.0145005440049701

-------

I took a look and it looks like one of the samples in my dataset now no longer has any data stored in PeakMatrix/chrX/jValues. Really unsure why it's now saying that it can't find these values for chrX in the same Arrow files now... I don't know what I could've done to somehow erase data from just chrX from just one of the samples in my dataset. Any thoughts on how this happened, so as to prevent this from happening in the future?

I'm attaching the log files for both the successful and the failed runs.

The log file for the successful run is named ArchR-getMarkerFeatures-67a13276200e-Date-2022-01-12_Time-20-21-30.log. The log file for the failed run is named ArchR-getMarkerFeatures-6a663d3298d0-Date-2022-01-12_Time-20-34-31.log.

Thanks for your help!
ArchR-getMarkerFeatures-67a13276200e-Date-2022-01-12_Time-20-21-30.log
ArchR-getMarkerFeatures-6a663d3298d0-Date-2022-01-12_Time-20-34-31.log