Figure 5
From: Classifying short genomic fragments from novel lineages using composition and homology
Percentage of classified query fragments assigned to the correct rank, correct lineage, or incorrectly. Each set of bars indicates the performance at a given rank when the child lineages of that rank are excluded from the training set. For example, results at the genus level are calculated with species level lineages excluded. Performance is reported at species (S), genus (G), family (F), order (O), class (C), phylum (P), and domain (D) ranks. The rank-specific NB-BL classifier always classifies query fragments at the strain level and as a result never assigns fragments to the correct rank. Results are reported for the 200 bp simulated test set. BLASTN and LCA results are for an E-value threshold of 10-5. The LCA classifiers use p = 15% and the ε-NB classifier uses ε = 105.