Modification calling missing for the inter-chromosomal translocated reads

nanoporetech / dorado

Oxford Nanopore's Basecaller

https://nanoporetech.com/

Other

439 stars 53 forks source link

Modification calling missing for the inter-chromosomal translocated reads #868

Open DexinYang1998 opened 3 weeks ago

DexinYang1998 commented 3 weeks ago

Hi,

First, I appreciate your wonderful tools for analyzing nanopore data. Recently, I have been trying to use Dorado 0.7.0 v5.0.0 HAC to call the base and modification from pod5 files, and I mapped the reads to the genome (chm13v2.0). I loaded the bam files in IGV and found that those WT reads (without any SV events) showed the CG methylation signal but those translocated reads with SV events seem to lose all the methylation signal.

base&modification calling and mapping are followed by this command using default parameters:

dorado basecaller <model> <reads> --reference <index> > calls.bam

IGV view:

WT reads:

SV reads:

Could you please help me figure it out? Thank you very much!

Dexin

DexinYang1998 commented 3 weeks ago

I am wondering whether the above issue is caused by the supplementary alignment.

malton-ont commented 3 weeks ago

Hi @DexinYang1998,

Yes, you are correct - MM, ML and MN tags are currently stripped from records that are marked as supplementary or secondary alignments unless soft clipping is enabled.

DexinYang1998 commented 3 weeks ago

Hi @DexinYang1998,

Yes, you are correct - MM, ML and MN tags are currently stripped from records that are marked as supplementary or secondary alignments unless soft clipping is enabled.

Thanks for your reply! I am wondering whether there is a way to solve this issue. Or do you think the issue is more relevant to the mapping software (minimap2) rather than Dorado?

malton-ont commented 3 weeks ago

Hi @DexinYang1998,

This is something done by dorado, not minimap2. I believe we had issues with trimming these tags correctly with hard clipping, and we took the decision that no data was better than bad data - @tijyojwad may have more information.

In the meantime, you could turn on soft clipping using the -Y argument.

oneillkza commented 3 weeks ago

@malton-ont when was hard-clipping made the default? I'm not seeing anything in the release notes. Going back to samples we ran on dorado 0.3.4, they still used soft clipping, but 0.6.1 uses hard.

This would have been useful to know about!

tijyojwad commented 3 weeks ago

Hi @DexinYang1998 - support for hard alignments is something we have in our backlog but we never got around to implementing it. We'll take a look at supporting it now.

@oneillkza - sorry we missed that in our CHANGELOG! we mentioned the dropping of modbase tags in the 0.5.1 release notes when hard clipping is enabled. The switch to hard clip by default happened in 0.5.0 from what I can see.

oneillkza commented 3 weeks ago

@tijyojwad would it make sense to change the default mode back to soft-clipping? It sounds like changing to hard-clipping has broken methylation all over the place.

e.g. modkit has issues with it too.

What was the rationale for defaulting to hard-clipping, anyway?

malton-ont commented 3 weeks ago

@oneillkza, the rationale was for the set of options for dorado alignment to be a subset of those for minimap2, which takes the same -Y parameter for soft-clipping. This came out of our review of the parameters as part of https://github.com/nanoporetech/dorado/issues/196.

FYI, I think this change was actually in 0.4.0 rather than 0.5.0.