Fig. 3

Benchmarking results of improving assemblies with LongStitch or LRScaf. For three human individuals, NA12878, NA19240 and NA24385, the wall-clock time and peak memory is shown for improving short-read ABySS assemblies and long-read Shasta assemblies using the default steps of LongStitch (up to ntLink) (orange) or LRScaf (blue). The peak memory (in gigabytes) is shown on a linear scale, and the wall-clock time (in hours) is shown on a log scale. All OPERA-LG runs on human data required over 9 days and 100 GB of RAM to complete, and are therefore not included in the plots