Do you want to perform this on the entire NCBI-nr database? Please note that it could pose significant scalability challenges, as the current NCBI-nr contains around 812 million accessions. Or are you considering working with just a subset of the database (to test)?