Hi everyone, I wanted to ask whether the pipeline's going to add

sourmash

for both the classification and the standardize steps in the future? It's a very useful tool so would love to have it in the pipeline 😁 Thanks!

It's on the roadmap definitely! We had an initial implementation like 2 years ago but unfortunately it didn't get finished. We recently added it to #C062H82UCRG, and @moritz in principle is championing this but he's on parental leave atm

Worse, I'm back to work 🥶

Nooooooooo

Related to this topic: We have sourmash in #C01U1E95JQ5, and found out that the naive approach to

sourmash gather

by just passing all samples and indices didn't find all matching genomes. For this reason, we've resorted to call the program once for each combination of sample and index and then filtering the collected output channel for unique genome identifiers. There might have been something we missed in how we call the module, but we do use the nf-core module.

Oof, that's good to know

Hi! Just adding to this it seems that especially for long-reads

sourmash

performs really well. Ive found this paper long time ago showing some benchmarks and you can look at the code implementation in this repo. Have you guys tested out how

sourmash

generally performs for metagenomics?

I've not tried it myself, but a lot of people appear to be quite happy like you say

yeah i'm doing sketch then gather and then tax (annotate)