nanoporetech / dorado

Oxford Nanopore's Basecaller
https://nanoporetech.com/
Other
493 stars 59 forks source link

Question: Inline basecalling with alignment #816

Closed damioresegun closed 4 months ago

damioresegun commented 4 months ago

Hi, I'm working on cDNA reads and I'm considering not using a reference during the basecalling and just outputting an unaligned BAM file. Am I correct in thinking that the unaligned BAM file will still have the right BC tag that demux needs for demultiplexing?

I'm also trying to understand what the benefit of using the reference during the basecalling is. Does it use the reference to act as a guide or is it just the the reference is used after basecalling is done and its a 'ease-of-use' thing to immediately optionally just get aligned reads out and it saves a step on the user's side?

Further to this, is there a way to add some more options to the minimap2 implementation for example adding --junc-bed and an associated bed file to the basecaller command?

Thanks!

tijyojwad commented 4 months ago

Am I correct in thinking that the unaligned BAM file will still have the right BC tag that demux needs for demultiplexing?

yes correct, demultiplexing is independent of alignment. barcodes are also trimmed by default when demultiplexing, so it's always better to demux before alignment (this is the sequence when running dorado basecaller with demux and alignment)

is it just the the reference is used after basecalling is done and its a 'ease-of-use' thing to immediately optionally just get aligned reads out and it saves a step on the user's side

yes, this

is there a way to add some more options to the minimap2 implementation

yes, we are working on adding this

damioresegun commented 4 months ago

Okay great, thanks. That clears things up for me. So that means that there is really no difference (given the same options) if I choose to output an unaligned BAM at the basecalling step, then do demux (--no-trim) vs if I did align with the reference at the basecalling step

tijyojwad commented 4 months ago

yes correct. if you're demuxing before aligning, then you can trim barcodes too. either is fine.

damioresegun commented 4 months ago

I'm choosing to keep them for now because I want to use pychopper downstream so I want to make sure the primers aren't affected beforehand. Unless you think dorado trim can remove the barcodes without affecting primer sequences?

tijyojwad commented 4 months ago

makes sense - I believe primers will get trimmed when trimming barcodes

damioresegun commented 4 months ago

Thanks for your help!