Skip to main content

Table 3 MiniScrub reduces downstream assembly errors

From: De novo Nanopore read quality improvement using deep learning

 

MECAT Raw

MiniScrub + MECAT

Canu Raw

MiniScrub + Canu

% genome assembled

79.39%

99.86%

99.69%

99.71%

NGA50

242478

1053459

1055037

696460

LGA50

12

3

2

5

# of contigs

38

11

7

19

# mis-assembled contigs

28

5

2

2

# local mis-assemblies

209

4

5

3

# indels > 5 bp

1099

394

84

46

Runtime (hours)

2.5

9

80

9

  1. MiniScrub significantly improves assembly, tested with MECAT [32], increasing genome coverage and NGA50 while limiting LGA50, mis-assemblies, mismatches, and indels. Canu’s assembly had slightly reduced errors and misassemblies when reads were preprocessed with MiniScrub, but the assembly was more fractured, likely due in part to resolving large misassemblies and indels. Notably, Canu assembly of raw reads took about 3.5 days, while the MiniScrub+Canu pipeline took about 9 hours, likely due to a reduction in the amount of error correction needed in the latter situation. Results were evaluated using QUAST [33] Best performance numbers are shown in bold