Quake: quality-aware detection and correction of sequencing errors

Table 4 SOAPdenovo bee assembly

Assembly	Trimmed Only	Corrected	Removed	Contigs	N50	N90	Scaffolds	N50	N90	Reads
Uncorrected Corrected	146.0 M	-	12.9 M	312,414	2,383	198	90,201	37,138	9,960	167.3 M
SOAPdenovo Corrected	134.4 M	15.7 M	15.6 M	188,480	4,051	515	36,525	36,525	9,162	164.8 M
Quake	146.9 M	16.5 M	13.0 M	189,621	4,076	514	37,279	37,014	9,255	167.3 M

SOAPdenovo assemblies of Megachile rotundata 124 bp paired end reads. We trimmed the reads before correcting with SOAPdenovo, which greatly improved its performance on our experiments with simulated data. The 'Trimmed only' column includes reads trimmed before and during SOAPdenovo correction. Quake trims reads automatically during correction. Correcting the reads reduces the number of contigs and scaffolds, increases the contig sizes, and allows the assembler to include more reads. Quake corrects more reads than SOAPdenovo which results in a slightly better assembly.