Bug: No peaks detected in not-blacklisted regions when a blacklist is given to macs3 hmmratac

[MariaRig]

Hi,
I have run macs3 hmmratac with and without a blacklist and I have realised that even small blacklisted regions (~1kb long can lead to huge gaps in the peak calling, but only in some samples. Not giving the black list fixed the issue.

These gaps seem more frequent in samples with worse signal-to-noise ratio, but the peak calling is overall very good except for the gaps.

Also, hmmratac was run on bams already filtered for the blacklisted regions, so it was probably unnecessary to run it with the black list, but I thought you would want to know that this can happen.

Kind regards

[taoliu]

Thanks! These peaks were possibly incorrectly removed when the 'candidate region' for Hidden Markov Model decoding overlaps just a tiny bit of the 'blacklisted' region. Let me think how to change this behavior. Agree that we may just need to remove the reads from alignment file in the beginning of the whole process.

[taoliu]

Re-implement the feature to exclude reads located in blacklisted regions before EM and HMM steps. 5b08668

Will be included in the next release.