Rebuilding CDR3 is still very hard

snijderlab / stitch

Template-based assembly of proteomics short reads for de novo antibody sequencing and repertoire profiling

MIT License

22 stars 3 forks source link

Rebuilding CDR3 is still very hard #231

Open douweschulte opened 1 year ago

douweschulte commented 1 year ago

Here is a very nice example where a couple of problems are visible:

Too short overhang for the last two reads
Slightly misaligned reads, asking the question if a multiple sequence alignment is warranted
With this level of rebuilding would an additional step between TM and REC be needed, essentially being REC but then use the germline with the reconstituted CDR3 as template (quite similar to #208)

MengTingHe2023 commented 11 months ago

I've also noticed this issue; the CDR3 of the recombined sequence generated with Stitch always seems to be much longer than the reference. And raising the CutoffScore doesn't seem to resolve the issue either.

douweschulte commented 11 months ago

Can you show me a picture of how your recombination looks (with the peptides visible for both sides IGHV + IGHJ) that way I can help you think if there is a way to get Stitch to recombine it better?