Fig. 2

Assembly evaluations. a Rank matrix for 40 Illumina assemblies. Columns correspond to metrics and are organized by metric classes. Rows correspond to assemblies and are sorted by assembler. Multiple assemblies were generated for each assembler differing by the input reads, parameters used, or both. Assembly nicknames correspond to Supplemental Tables S3 and S4. Assembly ranks span from worst (blue) to best (red). The red star marks the Platanus assembly used for hybrid assemblies. b-d Use the short-read assembly color scheme from (a) and the long-read color scheme from (e) to visualize (b) percent of complete BUSCOs found, c Log10 NG50, and d ALE scores. b and d show long-read scores before and after polishing steps. Dotted lines in (b-d) represent the best short-read assembly. e Rank matrices for 50 long-read assemblies organized as described in (a). Columns in the left matrix correspond to individual metrics whereas columns in the right matrix correspond to mean ranks of 40 different combinations of the 27 metrics. Red and blue stars mark assemblies used for BioNano scaffolding. Red stars represent the scaffolded assemblies that were selected for gene and repeat annotation. f-g Box and whisker plots of within-assembler rank distributions comparing blended (red) to PacBio-only (blue) inputs. Boxplots are not comparable between assemblers. Boxes show the 25th–75th percentile, the black line is the median, and the whiskers span the range. Assemblies from a given assembler were ranked either using (f) all individual metrics or (g) all ranked mean ranks from different combinations of metrics (see left and right panels of e). Ranks were then partitioned into those from blended versus PacBio-only assemblies. In both cases (f-g), blended assemblies from all assemblers except SMARTdenovo had significantly higher ranks by Wilcoxon Rank Sum Test than PacBio-only assemblies from the same assembler