Toward almost closed genomes with GapFiller

Table 1 Gap closure results obtained on the bacterial datasets

Method	Original	IMAGE	SOAPdenovo	GapFiller	GapFiller-LC
Escherichia coli
Genome size (bp)	4,478,287	4,530,961	4,490,973	4,490,638
Scaffolds	179	179	179	179
Gap count	544	291	16	11
Total gap length (bp)	12,516	2,861	16	130
Errors (SNPs)	12	40	33	22
Errors (indels)	4	17	25	9
Errors (misjoins)	1	1	1	1
N50	50,557	50,558	50,558	50,558
Streptomyces coelicolor
Genome size (bp)	8,558,275	8,576,331	8,557,720	8,558,333
Scaffolds	115	115	115	115
Gap count	158	63	60	23
Total gap length (bp)	9,221	4,009	1,288	806
Errors (SNPs)	299	423	406	280
Errors (indels)	664	677	769	686
Errors (misjoins)	12	17	18	18
N50	173,822	173,822	173,822	173,822
Staphylococcus aureus
Genome size (bp)	2,880,676		2,880,926	2,881,756	2,883,448
Scaffolds	19		19	19	19
Gap count	48		27	27	22
Total gap length (bp)	9,900		1,547	5,508	1,861
Errors (SNPs)	79		260	98	173
Errors (indels)	16		53	26	37
Errors (misjoins)	4		13	7	5
N50	1,091,731		1,091,333	1,092,281	1,092,421
Rhodobacter sphaeroides
Genome size (bp)	4,609,785		4,609,466	4,609,596	4,610,796
Scaffolds	38		38	38	38
Gap count	170		163	161	139
Total gap length (bp)	21,409		14,166	20,667	17,625
Errors (SNPs)	218		410	230	300
Errors (indels)	187		294	190	199
Errors (misjoins)	6		10	6	7
N50	3,192,334		3,192,075	3,192,215	3,192,974

Gap closure results obtained on four bacterial datasets show that the GapFiller strategy yields the most accurate finished genomes. Also, the gap count is lower compared to the other methods. The IMAGE method significantly underperforms on all quality measures and would therefore not be the preferred method to use. Differences are smaller between GapFiller and SOAPdenovo. Interestingly, whereas the gap count after closure is generally less for GapFiller, SOAPdenovo yields in three cases a shorter total gap length. This suggests the latter method is able to close larger gaps. Strikingly, however, the amount of errors is significantly higher for SOAPdenovo regardless of the source (SNPs, indels and misjoins). Even when applying less strict settings for GapFiller (GapFiller-LC: minimum coverage o = 1, ratio r = 0.5) to shorten the total gap length, our method still yields significantly less errors.

ISSN: 1474-760X