Fig. 2

Comparison of Metalign with competing methods on the challenge datasets from the first CAMI competition with organisms below 0.01% relative abundance excluded. Heatmaps show precision, recall, F1 score, and L1 error re-scaled such that higher is better (1 − (L1 error/2)) at the a family level, c genus level, and e species level. Scatter plots show Precision (x-axis) versus recall (y-axis) at the b family level, d genus level, and f species level. Metrics were averaged across all eight datasets