discrepancy between MethylDackel and IGV PBAT-style data aligned with bwameth about methyldackel HOT 11 OPEN

At the moment I'm not sure that's possible (for what it's worth, PBAT was just coming out when this package was created). Can you point to any convenient (ideally smallish, but I'll take big and high quality over small and low quality) public PBAT datasets? I can then play around a bit and see what works reasonably (or add an option to MethylDackel if needed).

from methyldackel.

Here is an example file where the image above comes from:

https://dl.dnanex.us/F/D/bygGZ2v3Y8qq8Pp8YGFyKyp81bp4xbJ67Bzk618X/CEG60-53-C1-oxBS_S1_L001_R1_001.1.cutB.seqtk_revcom.u.bwameth.paired.primary.barcoded.2Df.sorted.bam
https://dl.dnanex.us/F/D/kB19gVB0qjpP236Jzv9KjxBgjk7g7XgB0PqqQkPv/CEG60-53-C1-oxBS_S1_L001_R1_001.1.cutB.seqtk_revcom.u.bwameth.paired.primary.barcoded.2Df.sorted.bam.bai

It's small in size but has high coverage concentrated in certain regions.

from methyldackel.

Wonderful, thanks!

from methyldackel.

The reference is the standard GRCh38Decoy from NCBI:
https://dl.dnanex.us/F/D/jzGpFXgf85p8XQB3JVYY9fxYkFXVgQqP062Jx8v4/Homo_sapiens_NCBI_GRCh38Decoy.genome.fa

The reads where aligned with bwa-meth:

python bwa-meth/bwameth.py --threads 36 --reference include_indexes_dir/input/GRCh38Decoy.fa CEG60-53-C1-oxBS_S1_L001_R1_001.1.cutB.seqtk_revcom.u.fq.gz CEG60-53-C1-oxBS_S1_L001_R1_001.2.cutB.seqtk_revcom.u.fq.gz --prefix CEG60-53-C1-oxBS_S1_L001_R1_001.1.cutB.seqtk_revcom.u.bwameth

from methyldackel.

Let me know if I can supply any more relevant details...

from methyldackel.

Will do. I'm creating the indices at the moment.

from methyldackel.

A small update, I have a modified version of bwameth with a --pbat option that seems to be producing good results (I'm testing with DRR019425). I've only tested paired-end reads so far and need to give single-end reads a go. Once those are working nicely too I'll make a pull-request on bwameth, at which point you should be able to get some reasonable results with it. MethylDackel itself doesn't need any modifications, you can give it the BAM file from bismark (after sorting) and if bwameth with --pbat is used the results should be pretty similar (I get correlation coefficients of ~90%, with bwameth producing a LOT more alignments in a fraction of the time).

from methyldackel.

from methyldackel.

Please try the pbat branch with the --pbat option from my fork here. That now seems to be working correctly for SE and PE data.

from methyldackel.

@avilella Did you have a chance to test that bwameth branch? If so, did it produce reasonable results?

from methyldackel.

from methyldackel.