| L1 error | Precision | Recall/Sensitivity | F1-Score |
---|
MiCoP | 0.00909 | 1.0 | 1.0 | 1.0 |
Kraken | 1.15466* | 0.82222 | 0.925 | 0.87059 |
MetaPhlAn2 | 0.09844 | 1.0 | 1.0 | 1.0 |
- These species were sampled from the intersect of the species detected by all three tools in the previous simulation. Thus, this simulation consisted of only the species that were present in all three reference databases, eliminating reference bias. MetaPhlAn’s performance dramatically improved, predicting the exact set of species in the sample, but its abundance estimation was an order of magnitude worse than MiCoP’s. Kraken’s results did not markedly improve in this simulation. * Based on read classification proportions; Kraken does not claim to perform abundance estimation