Hi,
I am currently helping some research teams to optimize their current workflow which involves kneaddata and humann running on our local HPC. I recently discovered the nvBio / nvBowtie project and I wondered if kneaddata could be setup to use nvBowtie instead of bowtie2. This may help us achieve even greater speed as we have very nice GPU that could potentially crunch data much faster than our CPU.
Any idea if this is doable? At first I wanted to create a Singularity container, install kneaddata in compile nvBio in it and remove the bowtie2 binaries to replace them with the nvBowtie ones but I guess it might too naive 
Kind regards