Pseudo-Sanger sequencing: massively parallel production of long and near error-free reads using NGS technology

BMC Genomics

Table 1 Statistics of contigs assembled from simulated data

Dataset ^a	Program	Total length (bp)	Mean (bp)	N50 (bp)	N90 (bp)	Error ^b
D. melanogaster (simulation)	anytag ^c	113,166,478	66,141	197,693	43,974	109
	ABySS ^d	116,966,148	5,795	177,493	33,254	89
	MSR-CA	116,924,670	48,396	163,131	34,562	346
	soap ^d	113,971,825	16,208	56,061	13,361	90
	velvet ^d	114,719,611	16,573	104,879	23,729	330
Human Chr1 (simulation)	anytag	216,049,114	49,360	106,803	27,723	189
	ABySS	221,070,068	1,578	9,362	1,332	122
	MSR-CA	218,489,997	16,398	37,472	9,204	1,785
	soap	221,093,414	4,002	21,237	5,295	46
	velvet	Out of memory ^e

^aAll programs use the same simulated raw data. Our dataset was generated into four libraries, with insert sizes at 200 bp, 300 bp, 400 bp and 600 bp. Sequencing error was simulated at 0.005 and randomly distributed on the reads. The diploid heterozygosity is set at 0.001.
^bError = Inversion + Relocation + Translocation. The evaluation was completed by the evaluator from GAGE.
^canytag constructed pseudo-Sanger sequences, Newbler and minimus2 were used to assemble pseudo-Sanger sequences.
^dkmer size was iteratively set to 21, 25, 31, 41, 51 for ABySS, SOAPdenovo and velvet. The assembly with the largest N50 contig was showed.
^eOur memory limit is 450 G bytes.

ISSN: 1471-2164