CLARK: fast and accurate classification of metagenomic and genomic sequences using discriminative k-mers

Ounit, Rachid; Wanamaker, Steve; Close, Timothy J; Lonardi, Stefano

doi:10.1186/s12864-015-1419-2

BMC Genomics

Table 1 Genus-level classification accuracy and speed of CLARK, KRAKEN , and NBC for four simulated metagenomes and several k -mer length

From: CLARK: fast and accurate classification of metagenomic and genomic sequences using discriminative k-mers

	k	HiSeq			MiSeq			simBA-5			simHC.20.500
		*Prec*	*Sens*	*Speed*	*Prec*	*Sens*	*Speed*	*Prec*	*Sens*	*Speed*	*Prec*	*Sens*	*Speed*
NBC	15 ^∗	82.57	82.57	0.008	81.00	81.00	0.007	97.69	97.69	0.007	99.40	99.40	0.005
	13 ^∗	78.85	78.85	0.011	77.70	77.70	0.009	92.41	92.41	0.010	98.57	98.57	0.006
	11 ^∗	58.97	58.97	0.020	64.43	64.43	0.016	46.10	46.10	0.017	86.83	86.83	0.008
Clark(full)	31	99.26	77.78	541	95.33	77.69	435	98.88	89.67	591	99.68	99.42	121
	27	98.98	79.88	538	93.50	78.57	433	98.90	93.09	585	99.67	99.42	122
	23	97.33	81.97	530	90.06	80.02	426	98.71	94.54	559	99.59	99.42	119
	20	87.00	82.87	532	82.45	80.19	420	97.38	94.80	549	99.43	99.41	115
Kraken	31	99.26	77.76	2,332	95.50	77.59	1,361	98.28	89.35	1,976	96.83	96.55	237
	27	99.01	79.85	2,048	93.91	78.47	1,240	98.31	92.73	1,917	96.85	96.57	231
	23	97.45	81.89	1,923	90.56	79.75	1,186	98.25	94.18	1,824	96.80	96.57	228
	20	90.22	82.67	1,546	86.28	79.99	965	98.07	94.44	1,478	96.71	96.59	211
Clark	31	99.31	77.25	3,116	95.66	77.44	1,670	98.91	88.62	2,855	99.68	99.42	251
	27	99.07	79.37	2,796	93.90	78.29	1,522	98.90	92.26	2,554	99.67	99.42	241
	23	97.85	81.36	2,679	90.98	79.57	1,482	98.75	94.26	2,394	99.60	99.42	244
	20	88.60	82.26	2,567	83.35	79.77	1,456	97.73	94.49	2,306	99.43	99.41	239
Kraken-Q	31	99.20	76.84	6,224	95.81	74.13	5,308	98.17	87.46	7,023	91.17	85.79	3,809
	27	98.79	78.19	6,410	94.12	73.73	5,555	98.11	89.89	7,992	90.99	83.71	4,196
	23	96.67	78.48	7,015	90.57	72.35	6,329	97.21	89.07	8,989	90.46	79.27	4,574
	20	82.07	70.11	9,437	80.05	65.25	9,537	90.02	77.04	10,961	70.86	57.40	5,819
Clark-E	31	99.55	72.72	32,450	98.11	74.58	28,988	99.00	77.85	26,171	97.63	97.31	15,426
	27	99.43	74.67	29,897	96.93	75.68	28,459	98.93	84.86	27,451	97.47	97.18	16,124
	23	98.93	78.20	31,112	95.01	76.88	26,747	98.34	90.20	26,647	98.56	98.32	15,408
	20	94.74	78.46	30,029	90.57	76.60	25,789	96.61	89.98	26,545	93.94	93.82	15,587
Clark-l	27	98.45	62.30	1,525	92.11	69.64	861	95.96	52.00	1,705	99.49	98.94	143

Performance statistics for several choices of the k-mer length for NBC, KRAKEN, CLARK and their fast variants on the classification of “HiSeq”, “MiSeq”, “simBA-5” and “simHC.20.500” metagenomic datasets against the 695 genus-level targets; precision and sensitivity are expressed as percentages, while speed is expressed in 10³ reads per minute; KRAKEN-Q and CLARK-E are faster, but less accurate, variants of these tools; CLARK-l is a less memory-intensive version of CLARK which runs only for k = 27; experiments were carried out in single-threaded mode; ^∗parameter k is referred as N in the NBC manuscript.

Back to article page

ISSN: 1471-2164

Contact us

Submission enquiries: bmcgenomics@biomedcentral.com
General enquiries: ORSupport@springernature.com