Sébastien Rodrigue; Arne C. Materna; Sonia C. Timberlake; Matthew C. Blackburn; Rex R. Malmstrom; Eric J. Alm; Sallie W. Chisholm

“ In Figure 6A we show the fraction of DNA fragments for which MEGAN was able to assign to a known genome or taxon via significant protein homology. We show that composite Illumina reads from high-confidence alignments perform at least as well as reads obtained from 454-FLX. We also find that composite reads produced by alignments we designated as “low-confidence” have a very high fraction of unassignable reads, further proof of the ability of our algorithm to identify misalignments. ”

“ As the technology landscape evolves our general approach for experimental and computational overlapping of short paired reads will be readily adapted to new platforms and upgrades, and therefore is likely to continue to play a role in providing additional cost-effective sequencing options. Taken together, these results demonstrate that Illumina composite reads produced by SHERA constitute a practical and cost-effective alternative to 454-FLX for metagenome sequencing and analysis while providing significantly deeper sampling at a lower cost. ”

“ Composite read quality exceeds that of individual short-reads (Figure 3) . The shortest composite reads have the best quality scores, still, the average error rate remained <1% across 95% of 250 bp reads (Figure 5) . New technologies and sequencing chemistry improvements are likely to further increase ultra high-throughput read lengths, thus extending the achievable length of high-quality composite sequences.
Quality of Illumina reads out to 143 bp.A) The mean Phred quality of single reads is shown in the solid red lines, with error bars displaying quartiles. ”

“ The algorithm for aligning paired reads and computing quality scores was designed to be broadly applicable, independent of the sequencing platform. The source code is freely available at http://almlab.mit.edu/ALM/Software/Software.html. In brief, SHERA takes overlapping short reads as input, finds the best alignment, produces a composite read, calculates new quality scores for the overlapping bases, and scores the alignment confidence. ”

“ The size distribution of selectively bound DNA is concentration independent (Figure 2) , which supports the preparation of sequencing libraries from dilute samples. In the first step of this procedure, DNA fragments larger than the desired library insert size are bound to magnetic beads, which are then discarded. The supernatant is saved, and mixed with an appropriate volume of fresh SPRI beads. This second incubation preferentially immobilizes DNA fragments of the targeted size range. ”

“ For the alignment with the best score, the algorithm constructs the consensus sequence in this overlapping region by reporting the nucleotide with the higher quality value, or the informative base call if one nucleotide is 'N'. ”

“ This histogram is a sum of two distributions, the overlapper software's misalignments (a broad gaussian) and a sharp peak of small (1–2 bp) indels. We examined gapped alignments to confirm this was the case and also found that many of the insertion or deletions occur in homopolymer runs. We used a simple linear model to infer the number of misalignments ”