Arvados

I want to think about this one a bit. I think there's a couple ways to approach about this. One is tagging of individual files, which is what we've sketched out here.

Another another approach would be to design a more general table model for metadata -- a primary key (sample id / patient id or whatever) and a number of fields, which can include references to files.

As a data scheme, it needs to reasonably efficient for browsing in workbench, and basic filtering (ideally should be able to efficiently look up by any field).