The utax_split command splits a taxonomy reference database into two subsets as described in taxonomy training. The database must be in FASTA format with taxonomy annotations.
The level is specified by the -utax_level option. For example, -utax_level f specified family level. This means that a given family will be present in both output files but lower levels (genus, and species if present) will be present in one file or the other, but not both.
The output files are specified by the -output1 and -output2 options.
The split is generated randomly with a random number seed that varies in each run by default. To obtain reproducible results you can specify an integer random number seed using the -randseed option.
Example
usearch -utax_split 16s_ref.fa -split_level f -output1
split1.fa -output2 split2.fa -randseed 1