GAML: genome assembly by maximum likelihood

Table 2 Comparison of assembly accuracy in the first three scenarios

Assembler	Number of scaffolds	Longest scaffold (kb)	Longest scaffold corr. (kb)	N50 (kb)	Err.	N50 corr. (kb)	LAP
Staphylococus aureus, read sets SA1, SA2
GAML	28	1,191	1,191	514	0	514	−23.45
Allpaths-LG	12	1,435	1,435	1,092	0	1,092	−25.02
SOAPdenovo	99	518	518	332	0	332	−25.03
Velvet	45	958	532	762	17	126	−25.34
Bambus2	17	1,426	1,426	1,084	0	1,084	−25.73
MSR-CA	17	2,411	1,343	2,414	3	1,022	−26.26
ABySS	246	125	125	34	1	28	−29.43
Cons. Velvet*	219	95	95	31	0	31	−30.82
SGA	456	286	286	208	1	208	−31.80
Escherichia coli, read sets EC1, EC2
PacbioToCA	55	1,533	1,533	957	0	957	−33.86
GAML	29	1,283	1,283	653	0	653	−33.91
Cerulean	21	1,991	1,991	694	0	694	−34.18
AHA	54	477	477	213	5	194	−34.52
Cons. Velvet*	383	80	80	21	0	21	−36.02
Escherichia coli, read sets EC1, EC2, EC3
GAML	4	4,662	4,661	4,662	3	4,661	−60.38
Celera	19	4,635	2,085	4,635	19	2,085	−61.47
Cons. Velvet*	383	80	80	21	0	21	−72.03

For all assemblies, N50 values are based on the actual genome size. All misjoins were considered as errors and error-corrected values of N50 and contig sizes were obtained by breaking each contig at each error [24]. All assemblies except for GAML and conservative Velvet were obtained from [24] in the first experiment, and from [6] in the second experiment.
Italic numbers in each column signify the best result.
* Velvet with conservative settings used to create the assembly graph in our method.

ISSN: 1748-7188