Given a genome and a query, find characterized proteins whose descriptions match the query, and then search the genome for homologs of those proteins (example).
Or upload a genome or proteome in fasta format.
Curated BLAST relies on curated descriptions of over 100,000 experimentally-characterized proteins from Swiss-Prot, BRENDA, CAZy, CharProtDB, MetaCyc, EcoCyc, TCDB, REBASE, and the Fitness Browser.
Curated BLAST finds curated proteins whose descriptions match the query and uses ublast to compare these to the annotated proteins in the genome. Then, it uses ublast to compare the curated proteins to all possible proteins in the genome (the six-frame translation).
For details see the paper from 2019 or the code. Changes since the paper:
Lawrence Berkeley National Laboratory