Hi all, I've been working through this PR during the Hackathon I have been attending. Pretty much all of the code is there, except for the database setup code. Currently, it requires that you provide the databases yourself. I wanted to discuss how to go about doing this. For GPU padded databases, it's best that the administrator of the compute cluster setup the databases and then provide users access to them, given that GPU pads depend on the hardware. In my opinion, it would be best to provide documentation within the pipeline on how to set up padded DBs, rather than do it in the pipeline itself. Detecting hardware and computing databases accordingly could be quite tricky to do within the pipeline itself. It's a bit more complex than just downloading and untar-ing a database. That said, I have just had this nf-core module merged (nf-core/mmseqs/makepaddedseqdb) that can be used in the pipeline to compute the databases, if we wanted to go about implementing it into the pipeline. Curious on thoughts! 🙂

Hi Nathan, very nice! I tend to agree that might be tricky to maintain the creation of the padded DBs for different environments in the pipeline, but we may want to discuss this with others in the weekly development meeting, if we agree on this then the module should not be included in the PR, should it?

Hey @Jose Espinosa-Carrasco, thanks for getting back to me. I'm happy to discuss. I may not be able to make it as I have commitments on that day, but I will try pop in 🙂

OK, if you can not do it tomorrow feel free to join any other Tuesday, normally we have the meeting at 9:00 CET and I confirm in #C08THK11CHX channel the day before 😃