Comparison of classification methods using short read sequences while enforcing a statistical cutoff.

<p>Simulated read length is shown on the x-axis. In the top row, recovery is shown on the y-axis and refers to the proportion of queries with a correct taxonomic classification. In the middle row, erroneous recovery is shown on the y-axis and refers to the proportion of queries with an incorrect taxonomic classification. In the bottom row, coverage is shown on the y-axis and refers to the proportion of queries for which a classification could be made (correct or incorrect). The results for six taxonomic ranks are shown: kingdom (blue), phylum (red), class (green), order (purple), family (teal), and genus (orange). A ‘leave one out’ search approach was used with SAP. The asterisk indicates that NBC was run ‘as is’ from the Ribosomal Database Project website. Bars indicate standard error of the mean using four primers. The default statistical cutoffs for SAP (95% neighbor joining bootstrap proportion) and NBC (50% for sequences less than 250 bp, otherwise 80% confidence) are enforced.