Support bwa-meth alignment samples about methylbert HOT 6 CLOSED

Hello, thanks a lot for your interest!
Although I don't have first-hand experience with bwa-meth aligned BAM files, I am willing to update MethylBERT to support bwa-meth alignment in future. I just cannot guarantee when the update will be made.

Meanwhile, if you want to generate the XM tag yourself, you can find the information here (p.9 - 12):
https://www.bioinformatics.babraham.ac.uk/projects/bismark/Bismark_User_Guide.pdf

Or, if possible, I would rather recommend to generate a processed data file yourself including
"methyl_seq" column containing 0, 1, and 2 meaning unmethylated CpG, methylated CpG, and non-CpG. Please find the data format 03_Preprocessing_bulk_data.ipynb in the tutorial.
You don't need to make all of the columns. Only ["dna_seq", "methyl_seq", "ctype", "dmr_ctype", "dmr_label"] are needed in the file as shown in dataset.py.

Hope my comment is helpful for you!

from methylbert.

Thanks a lot! It's very helpful.

from methylbert.

Hello @vanreality

I just looked at the bwa-meth aligner github page but could not find information about how read-level methylation patterns are stored in bwa-meth-aligned bam files.

Could you please share it with me if you have any references/tutorials explaining read-level methylation patterns in bwa-meth?

Kind regards,
Yunhee

from methylbert.

As once mentioned in an issue in bwa-meth github page, the bwa-meth was designed as an atomic mapping tool, which means calling methylation and adding XM tag were left to other downstream tools.

I'm working on two ways trying to extract read-level methylation patterns from bwa-meth alignment.
Firstly I just found an R package epialleleR that seems applicable for adding XM tag. After testing, I'll let you know if it works well.
Next I'm still writing a homemade script for this. I'm not familiar with perl so I had a hard time reading bismark source code. (The commonly used methylation calling tool methyldackel could report read-level methylation rate, so its source code might also be worth reading)

from methylbert.

I evaluated the callMethylation function in the epialleleR package, version 1.11.9, using a WGBS sample. By comparing the XM tag similarity of identical reads (the same read name and position) between bwameth and Bismark, I found that 98.61% of the reads in the bwameth BAM files matched perfectly with those in the Bismark BAM files.

Given these results, I plan to use this R package to extract read-level methylation information from BAM files aligned with bwameth.

from methylbert.

Glad to hear this! I will then have a look at the epialleleR package :)
Thanks a lot!

from methylbert.