Fulgor: a fast and compact k-mer index for large-scale matching and color queries

Table 5 Total query time (elapsed time) and memory used during query (max. RSS) as reported by /usr/bin/time -v, using 16 processing threads

	Rate	Fulgor		Themisto-d1		Themisto-d20		MetaG.-B		MetaG.-NB		COBS
	Rate	m:ss	GB	m:ss	GB	m:ss	GB	mm:ss	GB	mm:ss	GB	mm:ss	GB
(a) low-hit
EC	4.71	0:10	1.65	0:33	3.22	0:33	2.35	7:34	2.82	3:40	0.38	10:25	28.94
SE-5K	1.27	0:09	0.77	0:32	2.27	0:30	1.77	6:48	2.76	2:55	0.31	11:50	37.64
SE-10K	13.86	0:10	2.01	0:36	5.32	0:36	4.06	7:35	3.00	4:17	0.56	14:33	75.63
SE-50K	32.61	0:25	17.91	1:05	37.45	0:56	33.07	8:33	5.05	6:47	2.42	39:33	367.34
SE-100K	34.09	0:45	41.49	1:39	81.60	1:22	75.89	9:19	7.04	7:33	4.23	48:52	\(^*\)521.58
SE-150K	34.01	1:06	69.05	5:02	130.94	2:05	124.19	—	—	—	—	37:40	\(^*\)522.47
GB	11.90	0:57	36.02	2:58	136.47	1:42	48.37	11:03	12.24	11:55	9.89	30:01	192.70

	Rate	Fulgor		Themisto-d1		Themisto-d20		MetaG.-B		MetaG.-NB		COBS
	Rate	mm:ss	GB	h:mm:ss	GB	h:mm:ss	GB	mm:ss	GB	h:mm:ss	GB	h:mm:ss	GB
(b) high-hit
EC	99.10	02:10	1.68	0:03:40	3.32	0:03:40	2.46	22:00	30.44	1:05:41	0.40	0:45:11	34.93
SE-5K	89.53	01:16	0.82	0:03:50	2.34	0:03:50	1.82	14:14	36:54	0:20:32	0.33	0:38:34	41.93
SE-10K	89.76	02:26	2.11	0:07:35	5.40	0:07:35	4.16	28:15	92.18	0:43:40	0.61	1:01:14	84.20
SE-50K	91.31	19:15	18.53	0:41:25	37.52	0:42:02	33.14	—	—	4:30:03	2.72	3:54:18	408.82
SE-100K	91.52	27:30	42.78	1:22:14	81.67	1:22:00	75.93	—	—	9:40:06	4.82	8:07:29	\(^*\)522.56
SE-150K	91.61	42:30	70.55	2:00:08	130.98	2:00:13	124.27	—	—	—	—	7:47:14	\(^*\)522.63
GB	92.98	01:10	30.02	0:02:45	136.55	0:01:20	48.47	28:55	15.86	0:22:05	9.91	0:34:45	225.57

The read-mapping output is written to /dev/null for this experiment. We also report the mapping rate in percentage (fraction of mapped read over the total number of queried reads). Results are relative to the full-intersection query mode (Algorithm 1). All reported timings are relative to a second run of the experiment, when the index is loaded faster from the disk cache. The “B” query mode of MetaGraph corresponds to the batch mode (with default batch size); and the “NB” corresponds to the non-batch query mode. With a \(^*\) we mark the workloads exceeding the available memory (\(>500\) GB). For the low-hit workload (a) we use the reads from SRR896663. For the high-hit workload (b) we use the reads from SRR1928200 for E. Coli, SRR801268 for S. Enterica, and ERR321482 for Gut Bacteria

ISSN: 1748-7188