search

LHentges / LanceOtron

https://LanceOtron.molbiol.ox.ac.uk/
GNU General Public License v3.0
22 stars 5 forks source link

Peak Scores when using LanceOtron #24

Open zhanglongnation opened 3 months ago

zhanglongnation commented 3 months ago

Hi, I have an H3K27ac CUT&Tag sequecing data for a non-model animal. A bw file was generated accoding to the protocol. After upload to the web, the process fail. Do you have any idea about the problem? I install the LanceOtron on my server, and it runs successfully. A bed file was generated, and the header contains many contents. Can you give a detailed explanation about the contents meaning? I noticed that overall_peak_score was in the generated bed file, is it resonable to exclued some peaks under a threashold? Many thanks

LHentges commented 3 months ago

My co-creator is looking into your issue using the webtool.

The overall_peak_score is the metric we published on in our paper, and is a combined assessment of both peak shape and multi-way enrichment. To answer your question, I think this is the best single score to filter on. In the paper we used a score of 0.5, which is equivalent to the model believing there is at least a 50% chance that the enriched region is derived from a biological signal rather than noise.

zhanglongnation commented 3 months ago

OK, I received the mail and it works now. Thanks a lot. Another question, peak calling using macs2 and Lanceotron may give different enrichment regions for the same region, do you have any suggestions for distinguishing which is more correct called peak? like the peak called in the attached file. LanceOtron give 4 peaks in the region, but macs2 gives almost 8 peaks, although it is more scatter. Peak_call_mcas2&LanceOtron.docx