SyMAP Results

Timing

Tested platforms

SyMAP has been tested on the following:

Machine	MySQL	Java	Core (CPU)	Memory	Purchased
v5.6.6 and later:
1. MacOS M4	MySQL v8.0.42	24 from Oracle	M4 12-Core	48Gb	2025
v5.4.1 and later:
2. MacOS x86_64	MySQL v8.4, MariaDB 11.0.2	8, 15, 17, 18, 20 from Adoptium and Oracle	3.2 GHz 6-Core	64Gb	2018
3. Linode (Ubuntu 22.04.2 LTS)	MySQL 8.0.33	17	Nanode	1Gb	2023
v5.4.0 and earlier:
4. Linux amd64 (Centos)	MariaDB v10.4.12	1.8	2.3 GHz 24-Core	128Gb	2011

Tested datasets and timings

Datasets

The following datasets have timings reported for them.

Mammalia	Brassicales	Poaceae	Rosales
Homo sapiens (Hsa)	Arabidopsis thaliana (A.thal)	Oryza Sativa (Rice)	Prunus persica (Peach)
Pan troglodtes (Pan)	Brassica rapa (B.rapa)	Zea Mays (Maize)	Prunus yedoensis (Pyedo)
Mus musculus (Mus)	Brassica oleracea (B.oler)	Triticum aestivum (Wheat)

MUMmer Timings

Most of the timings were on MacOS M4.

Times These are from the SyMAP symap.log file, which uses the Java system time functions (clock times are greater than the Java CPU system times). Times can vary over multiple executions; timings for only one execution is shown.

CPU Unless stated otherwise, the following #CPU was used: (1) 4 CPUs with amino acid alignment,
(2) 6 CPUs with self-synteny nucleotide alignment (NT). As mentioned earlier, it is important that the machine has at least 6G of RAM and disk space for the number of CPUs used; see CPU.

Memory SyMAP inputs chr x chr into MUMmer, so the largest chromosome size will have the biggest impact on memory. Concat concatenates small chromosomes; in the Notes column below, !Concat indicates this was not performed so that the input into MUMmer was smaller files; self-synteny never concatenates the chromosomes.

MacOS M4 with 48Gb has been tested with the following:

Species	MUMmer	Synteny	Size	Notes
	hr:min	min:sec
A.thal x B.oler	0h:31m	1m:14s	119Mb x 447Mb
B.raba x B.oler	1h:25m	4m:12s	297Mb x 447Mb	closely related
B.raba x self	0h:16m	1m:17s	297Mb x 297Mb	NT

Peach x Pyedo	0h:44m	1m:29s	227Mb x 408Mb	draft Pyedo ordered by peach

Hsa x Mus	30h:45m	5m:13s	3Gb x 2.7Gb	504 alignments, !Concat
Hsa x Mus	8h:03m	0m:37s	3Gb x 2.7Gb	MASKED, 84 alignments
Hsa x Pan	16h:40m	9m:23s	3Gb x 3Gb	MASKED, closely related, 84 alignments
Hsa x self	7h:45m	8m:28s	3Gb x 3Gb	276 alignments, NT

Wheat x Rice ^a	6hr:11m	3m:32s	14Gb x 373Mb	MASKED, CPU 1, largest chr 801Mb, !Concat
Maize x Rice	5h:59m	1m:16s	2.1Gb x 373Mb
Maize x self ^b	>48h	28m:52s	2.1Gb x 2.1Gb	duplications, NT
Maize x self	0h:16m	5m:03s	2.1Gb x 2.1Gb	MASKED, duplications, NT

Footnotes:

The largest wheat chromosome is 801Mb; in order for this to run on this relatively small machine:
(a) uncheck Concat, (b) both genomes Masked, (c) MUMmer V4 was used, (d) 1 CPU.
When CPU was set to 4, the machine hung because it ran out of application memory due to processing very large chromosomes in parallel.
The exact time for running MUMmer on Maize x self is approximate; there were 55 alignments that took anywhere between 1-5 hours each; I did not run it all at once. Maize has a lot of duplication, hence, the very long time (compare with Hsa x self).

MacOS x86_64 with 64Gb was tested on the following:

Species MUMmer SyMAP
A.thal x B.oler 0h:33m 4m:06s
Note the longer time for the synteny computation on the older Mac machine.
Linode nanode with 1Gb was too small to run MUMmer, so the MUMmer demo result files were transferred to the data/seq_results/demoArab_to_demoBrap directory. This allowed all other features to be tested on the demo, including running the synteny algorithm without the alignment. Also, two tiny input files were used to test MUMmer.
Linux amd64 with 128Gb was used extensively on large plant genomes.
For example, to align Maize x Rice used a total of 1h:3m using 8 CPUs

MUMmer MASK or Blast/Diamond

The following compares the datasets with (1) MUMmer non-masked, (2) MUMmer masked, (3) MUMmer masked with --maxmatch, (4) Blast, (5) Diamond. Diamond and Blast were run with blastp and -max_target_seqs 5, which produces the top 5 hits for each protein. The following were run on a MacOS M4.

Demo: Arabidopsis thaliana Chr1&5 x Brassica rapa Chr2&6

Size: 57Mb x 60Mb Genes: 13,476 x 8,138 CDSs: 70,655 & 41,428

MUMmer Params	Align (CPU 2)	Has Hits g2+g1+g0	Gene Pairs Hits	Gene Pairs Collinear	Gene Pairs Block	Avg %Id
Non-masked	5m:02s	8,506	7,701	3,396	5,787	67.0
Masked	3m:11s		7,549	3,274	5,518	68.1
--maxmatch	5m:15s		10,249	3,491	6,104	65.3
Blast Program -- Export: ~4min¹
Diamond	1s		30,522	4,645	7,906	53.7
Blast	4m:14s		64,744	4,960	8,108	42.2

¹Export Proteins: demoArab took 2m:53s and demoBrap took 1m:08s.

Brassica rapa (field mustard) x Brassica oleracea (wild cabbage)

Size: 297Mb x 447Mb Gene: 40,793 x 54,761 CDSs: 206,730 & 256,428

MUMmer Params	Align (CPU 4)	Has Hits g2+g1+g0	Gene Pairs Hits	Gene Pairs Collinear	Gene Pairs Block	Avg %Id
Non-Masked	1h:25m:10s	323,351	147,145	41,284	70,487	73.8
Masked	29m:05s		150,460	42,231	71,663	75.4
--maxmatch	52m:14s		420,995	51,637	82,441	68.3
Blast Program -- Export: ~24m¹
Diamond	15s		213,381	46,282	87,441	68.4
Blast	31m:27s		270,629	46,569	87,977	61.7

¹Export Proteins: B.rapa took 6m:50s and B.oleracea took 17m:14s.

Homo sapiens x Mus musculus

Size: 3.1Gb x 2.7Gb Gene: 19,888 x 22,133 CDSs: 189,229 & 207,696

MUMmer Params	Align (4 CPU)	Has Hits g2+g2+g0	Gene Pairs Hits	Gene Pairs Collinear	Gene Pairs Block	Avg %Id
Non-Masked	30h:45m:27s	41,058	21,304	14,611	15,302	63.3
Masked	8h:03m:16s		22,368	14,768	15,467	68.9
Blast Program -- Export: ~17min¹
Diamond	1s		65,665	21,208	24,660	58.2
Blast	19m:0s		109,291	22,036	26,308	48.0

¹Export Proteins: Humans took 8m:10s and Mouse took 9m:9s.

Note: Release v5.9.1 has a speedup for exporting proteins, which decreased the time for humans+mouse from ~2hr down to ~17min.

Go to top

Email: cas1@arizona.edu