See also
UTAX algorithm
utax command
Taxonomy benchmark home
Taxonomy benchmark tables
Taxonomy sensitivity vs. error plots
These are the methods tested in the taxonomy benchmark results.
UTAX_0.95, UTAX_0.90, UTAX_0.80, UTAX_0.60: UTAX with the given confidence cutoff. Version: usearch 8.1.1756.
RDP_100, RDP_80, RDP_50: The RDP Naive Bayesian Classifier with the given bootstrap cutoff. 100% is interesting for a benchmark test because it presumably gives the lowest possible error rate. 80% is the cutoff recommended by the authors for sequences longer than 250nt, 50% is recommended for sequences shorter than 250nt. The QIIME assign_taxonomy.py script uses a default cutoff of 50 regardless of length when the -m rdp option is used. Stand-alone classifier version: 2.11.
QIIME: the QIIME assign_taxonomy.py script with default options (uses the "uclust" method, which in fact is based on the USEARCH algorithm, not the UCLUST algorithm). QIIME version: 1.9.1.
QIIME-sortme: The QIIME assign_taxonomy.py script with the -m sortmerna option.
QIIME-BLAST: The QIIME assign_taxonomy.py script with the -m blast option.
mothur-knn: The mothur Classify.seqs command with the method=knn option. The mothur default is to use the 'wang' method which is its own implementation of the RDP Naive Bayesian Classifier, giving very similar results to the original. Mothur version: 1.33.3.
GAST: The GAST algorithm from Huse et al (2008). No version number. Source dated 25 Feb 2011.
SPINGO: v1.3 (32-bit). spingo -p 6 -i query.fa
-d db.fa > out.txt.