Comparison of available primary assemblers from long reads Benchmarking - Setonix HPC performance, using the flollowing settings on Slurm:
--exclusive - used exclusive node for benchmarking--cpus-per-task=24--nodes=1--ntasks=1--partition=work--mem=200G - maximum memory for work queue
Assembler
Wall-clock time
CPU time
Memory Utilized
CANU
04:52:07
4-15:32:44
62.21 GB
Flye
03:36:25
2-15:27:51
48.03 GB
NextDenovo
00:23:12
22:00:24
159.06 GB
Wtdbg2 (redbean)
00:33:29
12:27:42
6.19 GB
Raven
04:37:46
4-02:00:55
73.81 GB
Unicycler - lr
NA
NA
> 890 GB
Assembly contiguity and stats for test sample:
Assembler
# of contigs
N50
min
max
Genome size
CANU
53
2619752
1573
4510360
42.37e6
Flye
422
211792
574
1031083
38.88e6
NextDenovo
9
125536
82914
170627
1076432
Wtdbg2 (redbean)
209
595416
3277
1338536
36.82e6
Raven
177
689747
11657
4191860
45.38e6
Unicycler - lr
NA
NA
NA
NA
NA
GoldRush does not work with reads missing Phred quality scores, thus cannot be used for RSII reads PB.
Unicycler runs our of memory, thus not considered here for assembly.
Comparison for hybrid assemblers
Assembler
Wall-clock time
CPU time
Memory Utilized
Spades
03:45:11
2-22:34:13
18.24 GB
Unicycler - lr + sr
NA (time-out)
NA
NA
Assembly contiguity and stats for test sample:
Assembler
# of contigs
N50
min
max
Genome size
Spades
28688
342047
500
1121776
38.57e6
Unicycler - lr + sr
NA
NA
NA
NA
NA