Open AnneliektH opened 2 months ago
thanks for reporting! I'll look into it when I get a chance :)
side note - please try to put these in the branchwater plugin repo instead of the branchwater repo, please! I only found this because I keep these in my inbox until I get a chance to look at them :)
(you can also just put 'em in the sourmash issue tracker. doesn't need to be the branchwater plugin repo.)
Hey, I am running into some differences when running gather v fastgather, followed by tax metagenome. I suspect the issue is that fastgather is not using the --scaled parameter to do the calculations, but instead using the scale of the signature it is running on.
All was run in
/group/ctbrowngrp2/scratch/annie/2024-pigparadigm/results/sourmash
The comparisons were ran on the same metagenome, against gtdb-214-reps, with a scale of 1000 and a k 21.Gather/fastgather outputs are almost identical, except for the scaled value and thus also the total_weighted_hashes, which is 10x larger in the fastgather output
out:
Anyway, after running tax metagenome, there is a 10x difference at f_weigthed_at_rank. I think the one from gather is the correct one. I'll multiply the fastgather f_weigthed_at_rank by 10 for now, but I thought that maybe this is something that can be fixed: