modkit pileup for supplementary alignments?

[sklages]

pileup ignores (secondary) and supplementary alignments. In quite a few scenarios I'd like to see methylation events on both the primary and a (corresponding) supplementary alignment (e.g. translocation events).

This is nothing what I have been working with yet in great detail. So forgive me these somewhat very simple questions :-)

• is this an option to be considered for "pileup" in future? (-> including supplementary alignments)
• how can I "mimic" the pileup process atm? I could think about using "extract" and then afterwards aggregate all per-read counts to a per-position count, somehow?

Any ideas, hints?

Addendum: I use current ONT data, minimap2 or dorado aligner for alignment. modkit in current version 0.6.1.

[ArtRand]

Hello @sklages,

Supplementary alignments can be interpreted by Modkit if you perform basecalling and mapping with the appropriate options as described here (tl;dr pass -Y to minimap or dorado aligner). The reason that pileup does not use these records by default is that there is a risk of double-counting your reads, which can have an effect on global methylation level calculations as well as other gotchas. To be honest, I don't know why this can't be a non-default option (--allow-non-primary?). If I make a build that uses non-primary alignments in pileup, would you be willing to test it?

[sklages]

@ArtRand - sure :-) Thank you for the quick support..

[ArtRand]

@sklages Give this a shot.

$ modkit pileup \
secondary_and_supplementary_sorted.bam \
pileup.bed.gz \
--modified-bases 5mC 5hmC \
--allow-non-primary \ # <-- new flag
--bgzf \
--ref GCA_000001405.15_GRCh38_no_alt_analysis_set.fa \
--log modkit_pileup.log \
[--no-filtering]

Please send feedback if you have it. Also, try to run with the --log option and check it for anything that looks out of the ordinary.

modkit_v0.6.2_607e9ee.zip