Plassembler Simulated Reads Benchmarking Results

June 5, 2023 · View on GitHub

All benchmarking was conducted on a Intel® Core™ i7-10700K CPU @ 3.80GHz on a machine running Ubuntu 20.04.6 LTS.

The following Tables show the summary statistics of plassembler (with Flye and Raven) and Unicycler run from read sets simulated from 20 assemblies from Wick, Judd, Wyres et al 2021 here, Houtak et al (C222), here and here, Mathers et al (CAV1217) here, and for De Maio et al here.

Accuracy was assessed using QUAST comparing plasmid assemblies against the ground truth.

You can find the full pipeline used to generate these results here and the results themselves here.

Time and Memory Usage

As can be seen below, Plassembler was consistently faster than Unicycler by 3-10x in terms of wall clock time, with the largest increase coming single-threaded. Plassembler with Raven was also faster than with Flye.

ThreadsProgramMedian Wall Clock Time (s)Min Wall Clock Time (s)Max Wall Clock Time (s)Median Max Memory (MB)Min Max Memory (MB)Max Max Memory (MB)
1Plassembler with Flye7012192628103303924425275
1Plassembler with Raven389381522106176413072464
1Unicycler326191318366880280414043784
8Plassembler with Flye15635314852484222747712
8Plassembler with Raven7001323051257915877891
8Unicycler4411213910003656856106826
16Plassembler with Flye101949726755892383215086
16Plassembler with Raven43011417493517236015195
16Unicycler25541347509812967650913549

Assembly Accuracy

All results below were taken from each program run with 8 threads.

As can be seen below, plassembler missed only 1 plasmid (2370 bp plasmid in Enterobacter kobei MSB1 1B) while Unicycler missed 7, of which were small (under 10kbp). These were S. aureus C222 (2473 bp), C. koseri MINF 9D (9294 bp), K. oxytoca MSB1 2C (4574 bp), K. variicola INF345 (5783 bp), E. cloacae RBHSTW-00059 (2495 bp) and K. pneumoniae RBHSTW-00128 (3980 bp).

plassembler also had more fragmented assemblies than Unicycler (3 and 5 v 1). All assemblers fragmented the linear plasmid in K. variicola_ INF345, while Plassembler with Flye also fragmented a 3764 bp plasmid in C. gillenii RBHSTW-00142, and a 4841 bp plasmid in K. oxytoca RBHSTW-00167.

Brackets indicate (minimum and maximum). Indels and Mismatches are expressed as median, and per 100 kbp from QUAST.

ProgramMissed PlasmidsIncomplete/Fragmented PlasmidsMisassembliesGenome FractionIndelsMismatches
Plassembler with Flye13099.78 (mean), 99.97 (median), (98.16, 100)0 (0, 1.37)0.91 (0, 11.12)
Plassembler with Raven15099.04 (mean), 99.96 (median), (87.86, 100)0 (0, 1.37)1.04 (0, 11.12)
Unicycler71193.81 (mean), 99.88 (median), (0, 100)0 (0, 1.54)0.88 (0, 7.28)