I have the BLASTX results against nr database, I’m wondering if I can use HUMAnN 3 for doing functional analysis with this input? I tried to make the nr diamond database (using nr fasta file) but the generated outputs were empty.
That would work. You can read a bit more about working with custom protein databases here:
In your case (since you’ve already done the mapping to nr with blastx), you could create a separate id mapping file (as referenced in the text above) to map hits from nr sequences to the information HUMAnN needs for post-processing of results.
I’m not sure about the empty DIAMOND outputs. I assume you’re indexing nr protein with DIAMOND?