FindPeaks output

[DCarone]

Hi all,

For those of you using (or developing) FindPeaks for analyzing ChIPseq data - I have a question about how the score is calculated and how to interpret the score. Firstly, how exactly is the score calculated to come up with the numbers it does and can this be correlated to the number of sequences mapped to a given region? Also, if using the same parameters for 2 samples which I'd like to compare - say a control and a sample of interest, can I safely compare the scores to determine if there is enrichment in one sample versus another? Does the total number of sequences mapped have any influence on the score if they are different between the two datsets? I have tried both running each dataset with the same parameters and using my control dataset under the -control flag and then using -autothreshold and seem to get the same values so does this mean I can compare the peak heights to determine enrichment?

I'd appreciate any insight on analyzing the output data.

Thanks!

[ieuanclay]

Hi,

A very good paper came out recently : "PeakSeq enables systematic scoring of ChIP-seq experiments relative to controls" Rozowsky et al Nat. Biotech. (2009) 27(1):66-

It answers a lot of your questions, and also makes some very important points about normalising your data. Also about comparison of different datasets, with different numbers of reads in them.

Hope this helps,

Ieuan

[DCarone]

Thanks so much - I will check out the paper.

-Dawn

[ieuanclay]

No worries - even if you don't use their software, the lessons learned still apply.

Ieuan

[begsch]

'-control' option

Hello all,

I was wondering how FindPeaks 4.x.x does in detail the comparison of peaks in input data and control when using the '-control' option in order to estimate the likelihood that an input peak is part of the signal rather than noise? Is a linear regression-like best-fit line used for this task - as it's the case when the '-compare' option is applied - or is the algorithm more sophisticated?

Thanks in advance,
Ben