Bug: AssertionError when --cutoff-analysis was set

[freedog8]

Describe the bug
An exception was thrown when run callpeak program with --cutoff-analysis set.

Screenshots

Screenshots (no --cutoff-analysis)

**Screenshots (with control, maybe another issue) **

Another data with contorl

Another data with contorl (no --cutoff-analysis)

System (please complete the following information):

• OS: Linux
• python-3.7.5
• numpy-1.17.2
• macs2 2.2.5

[taoliu]

Hi @freedog8. In the first three cases, the fragment size inferred from BAMPE file is 0 and very few reads are "valid" after filtering. It seems that MACS2 can't correctly read the information. Could you share with me a chunk of your clean.bam file from bowtie? A way to get around this is to convert the BAMPE into a BEDPE file. In the BEDPE file, you just need three columns: chromosome, the leftmost position of the read pair, and the rightmost position of the read pair.

As for the "Another data with control", the complaint on rcond is caused by numpy. I will update the code to silence the warning. And the ValueError can be a quick fix. Stay tuned.

[taoliu]

hmm... I didn't realize that the find_optimal_cutoff function is an unfinished function from years ago... I have disabled that. The cutoff analysis should output a cutoff analysis table for you, but the 'suggest an optimal cutoff' feature should not be enabled.

The updates can be found in PR: #342 , and will be included in the next release.

[freedog8]

Hi @freedog8. In the first three cases, the fragment size inferred from BAMPE file is 0 and very few reads are "valid" after filtering. It seems that MACS2 can't correctly read the information. Could you share with me a chunk of your clean.bam file from bowtie? A way to get around this is to convert the BAMPE into a BEDPE file. In the BEDPE file, you just need three columns: chromosome, the leftmost position of the read pair, and the rightmost position of the read pair.

As for the "Another data with control", the complaint on rcond is caused by numpy. I will update the code to silence the warning. And the ValueError can be a quick fix. Stay tuned.

Oh, I just realize that the ChIP-Seq data of the first three cases is single-end. That‘s the reason. Thanks!

[taoliu]

Cutoff-analysis in callpeak cmd has been fixed in 2dd2839