A common approach of error correction of reads is to determine a threshold and correct k-mers whose multiplicities fall below the threshold (Chaisson and Pevzner, 2008; Kelley et al., 2010; Pevzner et al., 2001; Shi et al., 2010; Zhao et al., 2010).
This approach enabled the correction of reading frames of 23 protein genes, 25 pseudo genes, 21 rRNA genes and two tRNA genes (Additional file 2: Table S8).
After correction of reading frame errors, the original ORF encodes 247 amino acid residues with 81% identity to a type E neurotoxin (DELTA-BLAST alignment to residues 884 1,130 of most type E neurotoxins).
An optional position-specific correction of reads (step 6) can be performed if standard samples are provided by the user (typically a single cloned allele amplified and sequenced along with the samples; for details see Supplementary Materials).
This leads to the following trade-off in fragment assembly: either error correction of reads is performed, which may lead to errors in the contigs, or error correction is not performed and the complex de Bruijn graph has to undergo aggressive simplification that may lead to errors in the contigs.
The annotation of new genes and the correction of reading frames was based on transcriptional activity and was performed upon analysis of GC frame plots, ribosome-binding sites and -10 and -35 promoter regions using Artemis v12 [ 76] and comparisons to UniProtKB/Swiss-Prot, UniProtKB/TrEMBL, and InterProScan [ 70, 71].
Error correction of reads has thus come to the forefront as an essential problem, with a slew of publications in the last 2 years (Ilie et al., 2010; Kelley et al., 2010; Qu et al., 2009; Salmela, 2010; Schroder et al., 2009; Shi et al., 2010); Wijaya et al., 2009; Yang et al., 2010; Zhao et al., 2010;).
In reality, reads are error-prone necessitating error-correction of reads prior to any analysis.
For the evaluation of read correction quality, the metric gain has been established in (Yang et al., 2010, 2013) as a good summary of both sensitivity and precision.
Two recent software releases, from PacBio and the Celera Assembler group, allows for the error correction of PacBio reads without the need for short reads from another technology [ 16, 17].
