Table 1.
Comparison of the dbOTU implementations.
Table 2.
Benchmarks for the speed of the entire OTU calling process.
Fig 1.
Comparisons of communities analyzed by different methods.
dbOTU3 produces nearly identical results with dbOTU2 when visualized in a principal coordinate analysis ordination plot. Each point represents a community resulting from analysis of the mock community data one of the OTU callers. (The two triangles representing dbOTU2 and dbOTU3 always appear on top of one another, making a six-pointed triangle.) The “true composition” is the community composition expected based on how the communities were constructed. The principal components were computed using a matrix of the square roots of the Jensen-Shannon divergence between each pair of computed community compositions.
Table 3.
Benchmarks for the speed of the dissimilarity metric.
Fig 2.
Comparison of genetic dissimilarities.
The dbOTU2 metric (blue crosses) and dbOTU3’s Levenshtein metric (pink circles) predict true pairwise dissimilarities. Each point represents a comparison between a pair of sequences that were subjected to the genetic criterion while running dbOTU3 on the mock community data.
Table 4.
Accuracy of the genetic classification.
Table 5.
Correlation coefficients between the genetic dissimilarities.
Table 6.
Benchmarks for the speed of the distribution criteria.
Table 7.
Sequence count information for the case in which the simulated χ2 test and likelihood-ratio test disagree.
Table 8.
Likelihood-ratio test at varying p-value thresholds.