BMC Bioinformatics

Table 1 Choosing k and K

From: FMLRC: Hybrid long read error correction using an FM-index

	K
k	–	49	59	69	79	89
E.coli - Matching Bases
17	404736174	405150086	404809579	404608992	404361044	403888297
19	403117580	404580392	404571352	404418084	404297325	403826761
21	403002237	404365615	404367089	404255830	404131272	403841312
23	403516577	404381062	404378242	404240202	404107504	404041491
25	403819785	404461301	404480970	404453527	404363292	404236385
E. coli - Gain
17	0.1011	0.388	0.4709	0.5258	0.5468	0.521
19	0.3823	0.5887	0.612	0.6245	0.6279	0.6172
21	0.4879	0.634	0.6429	0.6459	0.6442	0.6345
23	0.5137	0.641	0.6474	0.6487	0.6457	0.6361
25	0.523	0.6396	0.6453	0.6461	0.6422	0.6318
S. cerevisiae - Matching Bases
17	1250679980	1253590990	1252340540	1251288299	1250445441	1249925285
19	1250052124	1252517259	1252462544	1252139063	1251853858	1251785285
21	1248322270	1251887685	1251963458	1251672602	1251758116	1251744201
23	1248801294	1252245368	1252387319	1252408890	1252545735	1252558864
25	1249574404	1252269051	1252478532	1252557840	1252778626	1252739127
S. cerevisiae - Gain
17	0.0264	0.224	0.3159	0.3946	0.452	0.4871
19	0.1172	0.3903	0.443	0.4822	0.5096	0.5273
21	0.2527	0.4938	0.5129	0.527	0.5367	0.5434
23	0.3319	0.5153	0.5251	0.5332	0.5388	0.5435
25	0.3728	0.5155	0.5226	0.5287	0.5334	0.5372

This table shows the result of running FMLRC using many different values for k and K for an E. coli and S. cerevisiae datasets
The test cases with K=− indicate that no second pass of correction using the long K-mer was performed, so those test cases use a single pass short k-mer only. After correcting the reads, we aligned the results using BLASR [22] and gathered statistics on the alignments. Matching bases indicates the number of matching bases across all mappings. Gain is defined as (TP−FP)/(TP+FN) (see “Correction accuracy” section). For each statistic, the best result is bolded in the above table. To summarize, increasing values for k and K tend to increase the gain but decrease the total matching bases - a tradeoff between sensitivity and specificity. Additionally, all tested values of K for a long K-mer pass improves the results over a single k-mer pass

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com