--no-model acceptable for peak calling?

[Chris Rhodes]

Hi I have some H3K27me3 ChIP-Seq data that is getting low numbers of peaks, possibly due to low signal-to-noise from sequencing. I have tried using default parameters on macs2 callpeaks as well as with the -broad option, but am getting a very low number of peaks (compared to a someone else's results). Looking at the stdout text, macs2 estimates my fragment size as 185. To try this with no modeling, I used "macs2 callpeak -t treatmetn -c control -g hs --broad --nomodel -extsize 185". Using the nomodel and the predicted fragment size from an earlier model-based macs run seemed to work fine. I got the expected number of peaks and it looks fine on UCSC.  Is using nomodel acceptable practice for calling peaks? Any suggestions for this approach?

Thanks,
Chris

[Chris Rhodes]

Found an answer under MACS announcement topic "using --nomodel option". It seems to cover my question well - no need for a reply.

Thanks,
Chris