Odd peak calling error

[Sara Knaack]

I have an error in MOSAiCS that I am looking for some expert advice on. I've successfully analyzed a set of ChIP-seq  data for Sfp1 in two yeast species, S paradoxus, and N. castellii. I analyzed these data with a specific pair of assemblies, applying the MOSAiCS fit in the "IO" mode, i.e. without mappability and GC content inputs; and successfully called peaks. I've repeated the alignment and binning of these same ChIP data using two different assemblies for these two yeast species. The data itself appears to be very similar in terms of the bin count frequency distribution with these new assemblies, and I apply the MOSAiCS fit in "IO" mode to all of the data in both species using these alternative assemblies without problem. In this test, for the N. castellii data I obtain complete and useful peak calling results after the MOSAiCS fit. For the S. paradoxus data, though, I now get an error for the peak calling step. I applied peak calling for this S. paradoxus data using the previous assembly, but now with only a different assembly for these same data, the peak calling function fails. The following error appears: 

Info: use two-signal-component model.

Info: calculating posterior probabilities...

  RCPP_FUNCTION_6 is deprecated, it will be removed permanently in july 2014

Info: calling peaks...

Error in peak_info_list[[1]] : subscript out of bounds

Calls: mosaicsPeak -> mosaicsPeak -> .local -> .peakCall -> matrix

Execution halted

The MOSAiCS fit provides a goodness of fit plot that looks quite reasonable for these same data with the different assembly, providing no clear indication why the peak calling fails. The error about the "subscript out of bounds" suggests to me no peaks may be found. I read in the data and apply the peak calling with the same arguments for both the N. castellii and S. paradoxus data with both sets of assemblies, (foundPeak <- mosaicsPeak(fit, signalModel="2S", FDR=0.05);). I've already applied peak calling on this data, just with a different assembly, so it is improbable no peaks would be found. I tried varying the FDR threshold value in the mosaicsPeak function between 0.75 and 0.01, but reproduced this error for each value. It also doesn't appear that any of the MOSAiCS model parameters are NA or non-reasonable values, which may also explain this. This second possibility seemed unlikely as the MOSAiCS plots seem reasonable. Finally, I also verified I was using the most up to date version of MOSAiCS from Bioconductor, and also for the needed dependency packages. 

I am hoping for recommendations for how to address such an error. Thank you very much for your attention!

Best regards,

Sara

[Dongjun Chung]

Hi Sara,

Thanks for your interest in MOSAiCS and I am sorry for the inconvenience due to the errors.

Regarding the error, I don't think that no peak generated errors because mosaics package can handle such cases (no peak genome-wide or no peak in any chromosomes).

If you don't mind, can you share your mosaics model fit you had the errors? Then, I will be able to figure out the sources of errors and fix them. Of course, this model fit will be handle confidentially.

Please just let me know whether this works for you.

Thanks,
Dongjun

--
You received this message because you are subscribed to the Google Groups "MOSAiCS User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mosaics_user_gr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.