Unsupervised statistical clustering of environmental shotgun sequences

Kislyuk, Andrey; Bhatnagar, Srijak; Dushoff, Jonathan; Weitz, Joshua S

doi:10.1186/1471-2105-10-316

BMC Bioinformatics

Table 5 The method of sampling the posterior distribution of the MCMC chain by averaging random accepted models from the steady state was compared to the method of selecting the model with the overall maximum log likelihood.

From: Unsupervised statistical clustering of environmental shotgun sequences

				Order 3 model			Order 4 model
Org 1	Org 2	Frag L	Sampling type	D ₃	Accuracy	LL	D ₄	Accuracy	LL
Arthrobacter aurescens TC1 vs. Sinorhizobium meliloti 1021
		400	Steady state sampled	1.08	0.95	-1054490.36
		1.09	0.94	-1040007.41
		400	Maximum log likelihood	1.02	0.94	-1055584.16
NC_003047	NC_008711	1000	Steady state sampled	1.95	0.97	-2648159.80
		2.52	0.99	-2637429.69
		1000	Maximum log likelihood	2.12	0.98	-2645204.57
Lactococcus lactis subsp. cremoris MG1363 vs. Francisella tularensis subsp. holarctica FTA
		400	Steady state sampled	1.08	0.90	-1045063.72
		1.33	0.95	-1040811.10
		400	Maximum log likelihood	1.15	0.92	-1047966.99
NC_009004	NC_009749	1000	Steady state sampled	2.02	0.96	-2624742.76
		2.22	0.97	-2615376.71
		1000	Maximum log likelihood	2.19	0.96	-2626080.18
Helicobacter pylori HPAG1 vs. Streptococcus pneumoniae R6
		400	Steady state sampled	0.93	0.96	-1059955.55
		1.18	0.93	-1045561.25
		400	Maximum log likelihood	0.97	0.96	-1061298.85
NC_003098	NC_008086	1000	Steady state sampled	1.71	0.99	-2656860.50
		2.28	0.99	-2634722.55
		1000	Maximum log likelihood	1.69	0.98	-2658488.27
Staphylococcus aureus RF122 vs. Prochlorococcus marinus str. NATL2A
		400	Steady state sampled	0.99	0.90	-1049716.33
		1.00	0.95	-1045188.54
		400	Maximum log likelihood	0.99	0.93	-1050316.80
NC_007335	NC_007622	1000	Steady state sampled	1.92	0.97	-2636903.64
		2.21	0.97	-2624299.41
		1000	Maximum log likelihood	1.75	0.97	-2636046.52
Staphylococcus aureus subsp. aureus COL vs. Methanocaldococcus jannaschii DSM 2661
		400	Steady state sampled	0.96	0.95	-1037936.55
		1.05	0.89	-1033285.36
		400	Maximum log likelihood	0.92	0.94	-1037505.67
NC_000909	NC_002951	1000	Steady state sampled	1.84	0.98
		2.36	0.99	-2581181.80
		1000	Maximum log likelihood	1.94	0.98	-2601394.32

Frag L, Fragment length; LL, Output model log likelihood
The resulting accuracy differences were negligible. Accuracy was also compared in 3-mer models vs. 4-mer models. While 4-mer models slightly outperformed 3-mer models on average, a significant run time increase was observed (not shown). NC _identifiers refer to GenBank accession numbers for genomes listed in each trial.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com