Peak Calling - Error when calculating statistical thresholds

[mw-neuro]

I have been trying to call peaks/generate narrow peaks files for a number of chip replicates and psuedo-replicates using the following settings:

-npeak=300000 -speak={Fragment_Length_Estimated_Previously_from_spp} -x=-500:{Max_Exclusion_Range_Estimated_From_First_100_Reads}

It has run without errors on about half of the files but for others I am getting the following error:

"" calculating statistical thresholds Error in if (!is.null(k) & dim(k)[1] > 0) { : argument is of length zero Calls: find.binding.positions -> lwcc.prediction -> do.call -> lapply -> FUN Execution halted ""

Based on a similar issue posted here where npeak files where not saved I have tried:

• changing the -fdr to 0.95
• removing the -npeak threshold
• specifying the npeak file with -savn='filename'
• a combination of all of the above.

But I still get the same error.

Any help on fixing this issue would be much appreciated!

[leepc12]

Anshul (author of the code) answered to your question. I am forwarding his comments.

It looks like SPP is not able to find any peaks. Looks like low quality datasets. Ask them to switch to using MACS2 if SPP is not finding any peaks. And ask them to use our pipeline instead of run_spp.R independently.

Please refer to our chip-seq-pipeline to call peak with MACS2 and then run IDR (phantompeakquals) on peaks (rep1 vs rep2 and repX-psedos for each repX ...). You can start from BAM or TAG-ALIGN.

{
    "chip.pipeline_type": "tf",
    "chip.paired_end": false, # or true if it's PE,
    "chip.nodup_bam": [ "REP1.filtered.bam", "REP2.filterered.bam"],
    "chip.ctl_nodup_bam": [ "CTL1.filtered.bam", "CTL2.filterered.bam"],
    "chip.peak_caller": "macs2", # or spp if you want to keep using it
     .... # other parameters
}

[mw-neuro]

Thanks very much for your reply! I will try calling peaks with MACS2