Limits...
Read length and repeat resolution: exploring prokaryote genomes using next-generation sequencing technologies.

Cahill MJ, Köser CU, Ross NE, Archer JA - PLoS ONE (2010)

Bottom Line: Nonetheless, there is considerable variation amongst prokaryotes.Given the diversity of prokaryote genomes, a sequencing strategy should be tailored to the organism under study.Our results will provide researchers with a practical resource to guide the selection of the appropriate read length.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, University of Cambridge, Cambridge, United Kingdom.

ABSTRACT

Background: There are a growing number of next-generation sequencing technologies. At present, the most cost-effective options also produce the shortest reads. However, even for prokaryotes, there is uncertainty concerning the utility of these technologies for the de novo assembly of complete genomes. This reflects an expectation that short reads will be unable to resolve small, but presumably abundant, repeats.

Methodology/principal findings: Using a simple model of repeat assembly, we develop and test a technique that, for any read length, can estimate the occurrence of unresolvable repeats in a genome, and thus predict the number of gaps that would need to be closed to produce a complete sequence. We apply this technique to 818 prokaryote genome sequences. This provides a quantitative assessment of the relative performance of various lengths. Notably, unpaired reads of only 150nt can reconstruct approximately 50% of the analysed genomes with fewer than 96 repeat-induced gaps. Nonetheless, there is considerable variation amongst prokaryotes. Some genomes can be assembled to near contiguity using very short reads while others require much longer reads.

Conclusions: Given the diversity of prokaryote genomes, a sequencing strategy should be tailored to the organism under study. Our results will provide researchers with a practical resource to guide the selection of the appropriate read length.

Show MeSH
Assessing the accuracy of the algorithm.The number of repeat-induced gaps predicted by the algorithm (grey bars) compared to the number of gaps observed (black bars) in actual assemblies of 36, 75, 125, 250, and 500nt simulated reads from A) M. genitalium, B) E. coli and C) S. coelicolor. The observed gaps are those between unique, non-redundant contigs larger than the read length. The coverage depth of each read set was the threshold at which random gaps are no longer predicted by the Lander-Waterman model. This occurs at effective coverage depths of 9–17×.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2902515&req=5

pone-0011518-g003: Assessing the accuracy of the algorithm.The number of repeat-induced gaps predicted by the algorithm (grey bars) compared to the number of gaps observed (black bars) in actual assemblies of 36, 75, 125, 250, and 500nt simulated reads from A) M. genitalium, B) E. coli and C) S. coelicolor. The observed gaps are those between unique, non-redundant contigs larger than the read length. The coverage depth of each read set was the threshold at which random gaps are no longer predicted by the Lander-Waterman model. This occurs at effective coverage depths of 9–17×.

Mentions: In Figure 3, the number of gaps between unique, error-free, contigs in each assembly is presented along with the predicted number of gaps based on the algorithm. If included, the error-containing contigs would have increased the total number of contigs by only 3.4% in the most extreme case.


Read length and repeat resolution: exploring prokaryote genomes using next-generation sequencing technologies.

Cahill MJ, Köser CU, Ross NE, Archer JA - PLoS ONE (2010)

Assessing the accuracy of the algorithm.The number of repeat-induced gaps predicted by the algorithm (grey bars) compared to the number of gaps observed (black bars) in actual assemblies of 36, 75, 125, 250, and 500nt simulated reads from A) M. genitalium, B) E. coli and C) S. coelicolor. The observed gaps are those between unique, non-redundant contigs larger than the read length. The coverage depth of each read set was the threshold at which random gaps are no longer predicted by the Lander-Waterman model. This occurs at effective coverage depths of 9–17×.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2902515&req=5

pone-0011518-g003: Assessing the accuracy of the algorithm.The number of repeat-induced gaps predicted by the algorithm (grey bars) compared to the number of gaps observed (black bars) in actual assemblies of 36, 75, 125, 250, and 500nt simulated reads from A) M. genitalium, B) E. coli and C) S. coelicolor. The observed gaps are those between unique, non-redundant contigs larger than the read length. The coverage depth of each read set was the threshold at which random gaps are no longer predicted by the Lander-Waterman model. This occurs at effective coverage depths of 9–17×.
Mentions: In Figure 3, the number of gaps between unique, error-free, contigs in each assembly is presented along with the predicted number of gaps based on the algorithm. If included, the error-containing contigs would have increased the total number of contigs by only 3.4% in the most extreme case.

Bottom Line: Nonetheless, there is considerable variation amongst prokaryotes.Given the diversity of prokaryote genomes, a sequencing strategy should be tailored to the organism under study.Our results will provide researchers with a practical resource to guide the selection of the appropriate read length.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, University of Cambridge, Cambridge, United Kingdom.

ABSTRACT

Background: There are a growing number of next-generation sequencing technologies. At present, the most cost-effective options also produce the shortest reads. However, even for prokaryotes, there is uncertainty concerning the utility of these technologies for the de novo assembly of complete genomes. This reflects an expectation that short reads will be unable to resolve small, but presumably abundant, repeats.

Methodology/principal findings: Using a simple model of repeat assembly, we develop and test a technique that, for any read length, can estimate the occurrence of unresolvable repeats in a genome, and thus predict the number of gaps that would need to be closed to produce a complete sequence. We apply this technique to 818 prokaryote genome sequences. This provides a quantitative assessment of the relative performance of various lengths. Notably, unpaired reads of only 150nt can reconstruct approximately 50% of the analysed genomes with fewer than 96 repeat-induced gaps. Nonetheless, there is considerable variation amongst prokaryotes. Some genomes can be assembled to near contiguity using very short reads while others require much longer reads.

Conclusions: Given the diversity of prokaryote genomes, a sequencing strategy should be tailored to the organism under study. Our results will provide researchers with a practical resource to guide the selection of the appropriate read length.

Show MeSH