Fig. 4

Curation of entire binning plans on the CAMI2 marine dataset. A and B The gold standard assembly was visualized using PCA on tetranucleotide frequencies, showing DASTool-binned contigs pre- (A) and post- (B) curation using BinaRena. Contigs are colored by the silhouette coefficient calculated by BinaRena. Marker size (radius) is proportional to the cube root of contig length. Marker opacity is proportional to the 4th power root of average contig coverage of 10 samples. C–G Metrics of three binning plans (generated by DASTool, MaxBin, and MetaBAT, respectively) pre- and post-curation. C Total number of contigs in bins. D Total number of bins. E Total length of contigs in bins. F Adjusted Rand index between each binning plan and the ground truth as calculated by BinaRena. G Numbers of high-, medium-, and low-quality MAGs, defined following [1] based on CheckM-inferred completeness and contamination scores. Specifically, high quality: ≥ 90% complete, < 5% contaminated; medium quality: ≥ 50% complete, < 10% contaminated; and low quality: < 50% complete, < 10% contaminated. Bins that do not match any catalog (i.e., ≥ 10% contamination) are excluded