run_spp.R with spp v1.11

[xi wang]

Hi everyone, 

Applying the script run_spp.R onto the latest version of spp (v1.11), an error always occurred, as follows:

Error in write.narrowpeak.binding(region.peaks, iparams$output.rpeak.file,  : 

  unused argument(s) (npeaks = iparams$npeak)

Execution halted

Seems that the newest version has removed the argument "npeaks". And my question is why we need to specify "npeaks" for IDR analysis? 

Thanks

Xi

[Anshul Kundaje]

I had modified the SPP package to allow it to run with the npeaks argument. The modified version is provided in my package. So please use the one provided. I will update it with the new version of SPP in the next update.

You dont have to use the npeaks arguments. The main point is that you want to call sufficient number of peaks that the IDR algorithm gets to see a sufficient noise (fake peaks) component. The npeaks parameter allows you to obtain a certain number of ranked peaks (if they exist) rather than selecting an FDR threshold. If you really want to use the new version of SPP, then use an fdr threshold of 0.9 and pick the top 150-200k peaks (if you are working with human or mouse or top 30k if you are working with worm or fly) from the output peak lists (if the number of peaks exceed this number) to feed to IDR.

Thanks,

-A

--