construct ROC from SPP/idr/conservative set

[Yan Zhu]

Hi,

I would like to construct ROC curve for the peaks that are called from SPP and filtered by idr (my peaks vs ENCODE gold standard peaks). In MACS case, I would use p-value, but SPP result didn't assign p-value. I would imagine to use Z-score for SPP peaks in this case for construing ROC curve, should I use signal value(col 7) or -log10(q-value) (col 9)?

Thanks

Yan

[Anshul Kundaje]

Use column 7 i.e. peak score from the narrowPeak files.

-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+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[Yan Zhu]

Thanks for your reply, does this signal value (column 7) represent statistical significance ?

and a more general question, why ENCODE suggests to use SPP for TF peak call but MACS for histone?

Thanks

Yan

On Wednesday, February 14, 2018 at 6:58:39 PM UTC-5, Anshul Kundaje wrote:

Use column 7 i.e. peak score from the narrowPeak files.

-Anshul.

On Wed, Feb 14, 2018 at 3:52 PM, Yan Zhu <yan...@vt.edu> wrote:

Hi,

I would like to construct ROC curve for the peaks that are called from SPP and filtered by idr (my peaks vs ENCODE gold standard peaks). In MACS case, I would use p-value, but SPP result didn't assign p-value. I would imagine to use Z-score for SPP peaks in this case for construing ROC curve, should I use signal value(col 7) or -log10(q-value) (col 9)?

Thanks

Yan

--
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.

[Anshul Kundaje]

No the signal value is ChIP - input with a penalty for peak shape (You'll have to look at the exact details in the SPP paper). So its a signal score.

SPP consistently exhibits better reproducibility than MACS2 but is primarily designed for punctate sharp binding events. So we use it for TFs. MACS2 does a better job with variable width enriched regions like those from histone marks.

Thanks,

Anshul.

To unsubscribe from this group and stop receiving emails from it, send an email to idr-discuss+unsubscribe@googlegroups.com.