i found https://github.com/bluenote-1577/savont it advertises itselfs as DADA2 but for long reads do you think that with this we might be able to integrate long reads into ampliseq?

generally that looks interesting. I miss some benchmarking (except the comparison with unoise there is nothing?).

it has some preeliminary results here https://github.com/bluenote-1577/savont/wiki/Savont-Preliminary-Results it has 3 bechmarks

havent seen that, looks indeed promising

Interesting tool. We’ve actually been working on a ‘rewrites’ version of DADA2 in Rust which seems to be going pretty well, but this would be a nice alternative for our HiFi runs.

Yes, the comparison is not entirely fair. But I dont think that the idea was to discredit UNOISE3 but rather to show that savont is better than existing solutions (that are known to be imperfect for ONT). Do you know any software that might be better for ONT amplicon -> ASV than UNOISE3, i.e that would be fairer to compare to savont?

I raised this question here: https://github.com/bluenote-1577/savont/issues/6, and Jim Shaw (bluenote) linked to a biorxiv preprint: https://www.biorxiv.org/content/10.64898/2026.02.26.708165v1. But he also pointed out he’s using a different approach