Fig. 2: Performance of genome binners on short-read assemblies (GSA and MA, MEGAHIT) of the marine, strain-madness, and plant-associated data.
From: Critical Assessment of Metagenome Interpretation: the second round of challenges
a, Boxplots of average completeness, purity, ARI, percentage of binned bp and fraction of genomes recovered with moderate or higher quality (>50% completeness, <10% contamination) across methods from each dataset (Methods). Arrows indicate the average. b–g, Boxplots of completeness per genome and purity per bin, and bar charts of ARI, binned bp and moderate or higher quality genomes recovered, by method, for each dataset: marine GSA (b), marine MA (c), strain-madness GSA (d), strain-madness MA (e), plant-associated GSA (f) and plant-associated MA (g). The submission with the highest F1-score per method on a dataset is shown (Supplementary Tables 9–15). Boxes in boxplots indicate the interquartile range of n results, the center line the median and arrows the average. Whiskers extend to 1.5 × interquartile range or to the maximum and minimum if there is no outlier. Outliers are results represented as points outside 1.5 × interquartile range above the upper quartile and below the lower quartile.
