IDR threshold - 0.1 or 0.05?

[Lucy]

Hi,

I was wondering whether it is preferable to use an IDR threshold of 0.1 or 0.05?  In the Kundaje protocol that I have (I'm not sure whether it is the most recent version), it says to use 0.05 for ENCODE and 0.1 otherwise.  I think my data is good quality, so in this case, is it better to be stringent, or what is the reason for being more relaxed (i.e. using 0.1)?

Many thanks for the help - sorry I'm new to all this!

Lucy

[Anshul Kundaje]

Yes use 0.05 as far as possible if your goal is to avoid potential false/irreproducible events.

If you want higher sensitivity for a specific application you can use 0.1 or use the naive-overlap peaks that the pipeline produces.

Btw I would recommend using our automated pipelines as far as possible unless you are a power user and have specific use cases.

https://github.com/kundajelab/chipseq_pipeline

https://github.com/kundajelab/atac_dnase_pipelines

-Anshul.

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

[laij...@gmail.com]

Hi! I was wondering how to apply an IDR threshold to the naive-overlap peaks. Which column should we use? When I use column 5 (scaled IDR value, 0.05 threshold corresponding to 540) there are 0 peaks. When I use column 11 (global IDR value? <0.05) then I get 90k peaks. Thank you for your help!

[Anshul Kundaje]

https://github.com/kundajelab/idr

Please read the file format carefully. 

Column 12 is globalIDR float -log10(Global IDR value)

So use that column to threshold. Note the values are in -log10 scale. 

Anshul

To view this discussion on the web visit https://groups.google.com/d/msgid/idr-discuss/f631a205-8960-4a42-85f8-e975794a76d3n%40googlegroups.com.