1. Download the 16S reference files and extract them from the archive using tar -zxvf. The archive contains a FASTA database with full-length sequences plus pre-trained taxconfs files for shorter segmenets (500nt, 250nt and 120nt).
2. Choose the taxconfs file for length closest to your sequences.
3. Make the UDB file using the makeudb_utax command (it is ok to use full-length reference sequences with parameters for shorter sequences):
usearch -makeudb_utax refdb.fa -output refdb.udb -taxconfsin 500.tc
4. Run the utax command or cluster_otus_utax command:
usearch -utax reads.fastq -db tax.udb -utaxout
utax.txt -strand both