Sequencing of BAC pools by different next generation sequencing platforms and strategies
- Stefan Taudien†1Email author,
- Burkhard Steuernagel†2,
- Ruvini Ariyadasa2,
- Daniela Schulte2,
- Thomas Schmutzer2,
- Marco Groth1,
- Marius Felder1,
- Andreas Petzold1,
- Uwe Scholz2,
- Klaus FX Mayer3,
- Nils Stein2 and
- Matthias Platzer1
© Taudien et al; licensee BioMed Central Ltd. 2011
Received: 22 September 2011
Accepted: 14 October 2011
Published: 14 October 2011
Next generation sequencing of BACs is a viable option for deciphering the sequence of even large and highly repetitive genomes. In order to optimize this strategy, we examined the influence of read length on the quality of Roche/454 sequence assemblies, to what extent Illumina/Solexa mate pairs (MPs) improve the assemblies by scaffolding and whether barcoding of BACs is dispensable.
Sequencing four BACs with both FLX and Titanium technologies revealed similar sequencing accuracy, but showed that the longer Titanium reads produce considerably less misassemblies and gaps. The 454 assemblies of 96 barcoded BACs were improved by scaffolding 79% of the total contig length with MPs from a non-barcoded library.
Assembly of the unmasked 454 sequences without separation by barcodes revealed chimeric contig formation to be a major problem, encompassing 47% of the total contig length. Masking the sequences reduced this fraction to 24%.
Optimal BAC pool sequencing should be based on the longest available reads, with barcoding essential for a comprehensive assessment of both repetitive and non-repetitive sequence information. When interest is restricted to non-repetitive regions and repeats are masked prior to assembly, barcoding is non-essential. In any case, the assemblies can be improved considerably by scaffolding with non-barcoded BAC pool MPs.
KeywordsBAC pools next generation sequencing 454 Illumina barcoding mate pairs scaffolding barley
With the establishment and widespread use of massively parallel next generation sequencing (NGS) platforms, de novo sequencing of large complex plant genomes is now feasible [1–3]. For such endeavours, a mixture of whole genome shotgun (WGS) and "clone-by-clone" sequencing is generally advised. While the first approach is based on random shearing of total genomic DNA, the second method relies on a pre-defined minimum tilling path (MTP) of large insert clones which are anchored to a genetic map ("hierarchical shotgun").
Due to its accuracy and reliability, the latter strategy is favourable for producing high-quality reference sequences such as the arabidopsis [4, 5], (http://www.maizegdb.org) and barley (Hordeum vulgare)  (http://barleygenome.org) genomes. Unfortunately, clone-by-clone sequencing is more costly and labour intensive than WGS. Additionally, the massive sequence data produced by a single NGS run (Roche/454 GS Titanium up to 400 Mb; Illumina/Solexa GAIIx up to 8 Gb/lane) requires pooling of BACs. This additional complexity increases with the number of pooled clones and can hamper de novo assembly, particularly with BACs harbouring high fractions of repetitive sequences, such as those derived from barley . To reduce these assembly challenges, clones can be selected by mapping information. Previous works have sequenced plant genome derived clone pools, both with and without complexity reduction. After a pilot experiment using four BACs from barley , Rounsley and co-workers  reported the combined 454 shotgun/PE sequencing of a 19.4 Mb rice (Oryza barthii) chromosome arm in six pools of 168 non-barcoded overlapping BACs. Additionally, the 454 sequencing of 91 barcoded non-overlapping barley BACs in pools of up to 24 clones without additional PE/MP information was described by our group . Recently, Gonzalez et al.  sequenced 58 non-barcoded, non-overlapping BACs from melon (Cucumis melo) in two pools by 454 shotgun/PE and supported by BES. These assemblies were partially checked for misalignments by comparison to high quality references such as Sanger sequenced clones [10, 11] or to the highly similar genome of O. sativa.
the 454 shotgun read lengths and sequence depths matter for the consistency and accuracy of the assemblies;
Illumina MP read informations improve the 454 shotgun assemblies by scaffolding;
assemblies based on barcoded 454 sequences are superior to those of non-barcoded clones.
Sequencing of reference BACs with 454 FLX and Titanium chemistry
To address the first question we used a set of four non-overlapping BACs (184G08, 259I16, 631P08, 711N16) previously sequenced with Sanger technology (AY268139, AF474373, DQ249273, AF427791)  ("reference BACs"). These BACs were recently sequenced as part of pools using the 454 FLX chemistry , with mean read lengths between 219 and 225 bp and clone sequence depths between 15x and 27x.
For the present study, the barcoded reference BACs were sequenced again with the 454 Titanium chemistry. The obtained sequences were separated according to the BAC specific barcodes and clipped for vector and barcoding motifs resulting in average read lengths of 252 to 292 bp. Alignment to the Sanger references revealed average sequence depths between 25x and 66x (additional files 1,2,3). For convenience, in the following the data from barcoded BACs obtained by FLX and Titanium sequencing are abbreviated as "bcFLX" and "bcTi", respectively.
Evaluation of the consistency of assemblies
Comparison of the 454 assemblies of barcoded BACs with their Sanger reference sequences.
average read length (bp)
total gap size (bp)
184G09 120,562 bp
259I16 124,050 bp
631P08 101,158 bp
711N16 112,178 bp
Better assemblies can result from longer reads, higher sequence depths or both. In general, the Titanium technology produces longer reads than the GS FLX platform. For the reference BACs we achieved median bcFLX read lengths of ~240 bp and 25/75% quartile lengths of ~230/255 bp with the upper 1.5x interquartile values at 300-310 bp. In contrast, the bcTi reads showed higher variation in length with upper 1.5x interquartile lengths of ~600 bp (additional file 5). In fact, these long sequences are likely the major cause for the better bcTi assemblies. On the other hand, the different mean sequence depths (47x for bcTi and 23x for bcFLX) may also influence this improvement, but the impact of both parameters on the assembly quality and consistency could not be independently evaluated in this data set. To overcome this restriction and estimate the influence of read length on the assembly qualities, we reduced the sequence depths of the bcTi datasets to the level of bcFLX. As input, 20 independently and randomly "down sampled" sequence sets were used for each of the four BACs and the resulting contigs were mapped to the Sanger reference in the same way as reported above. Comparison of the down sampled ("bcTids") to the corresponding bcTi assemblies showed different trends with respect to the number of misassemblies and gaps as well as to L50/L80 (additional files 4 and 6). For 184G09 and 711N16, bcTi and bcTids assemblies were equivalent. For 259I16 down sampling led to more misassemblies and more gaps as well as to shorter L50/L80 lengths compared to bcTi. Interestingly, for 631P08, down sampling reduced the number of misassemblies (1 instead of 2) as well as the number of gaps (1 instead of 2) together with an increase of L50 (52 vs 26 kb). To quantify the comparison of the bcTi/bcTids with the bcFLX assemblies, we defined penalties: 2 per misassembly, 1 per gap. As result, for all BACs, the bcTids penalties were smaller than the bcFLX ones (Table 1).
Estimation of sequencing accuracy
Error rates of different chemistries by comparison to the Sanger reference sequences.
Mate Pair sequences for scaffolding BAC assemblies from barcoded 454 sequences
We also investigated the utility of MP reads for 454 single-read assembly improvement by scaffolding. 48 barcoded BACs were sequenced using the FLX (pool 1) and Titanium chemistry (pool 2). The BAC-specific assemblies resulted in 1,473 contigs with a total length of ~11.1 Mb. Pool 3 contained all 96 BACs of pools 1 and 2 and was sequenced using a 3 kb MP library on the Illumina platform (additional files 10,11). After removal of duplicates, we obtained ~106 pairs of 2 × 36 bp, corresponding to ~82 Mb. Mapping pairs to the reference of one BAC (562B07) revealed a median distance of 2,825 bp (1.5 x interquartile range: 1,922..3,742 bp; additional file 12). MPs were mapped against the 454 assemblies and gap bridging MPs with correct orientation to each other and a total of distances to the contig ends up to 3,742 bp were extracted. Altogether, 1,665 contig pairs are bridged by 52,234 MPs with 1 to 561 MPs per link.
Scaffolding contigs from 454 assemblies of 96 BACs by Illumina MPs
gap bridgings, total
gap bridgings, discarded 1
gap bridgings, subjected to scaffolding
conflict free scaffolding
not scaffolded due to missing MPs or conflicts
Comparison of the 96 barcoded BAC 454 assemblies without and with scaffolding by Illumina MPs.
contigs and scaffolds
L50 1 [bp]
Comparison of assemblies with barcoded and non-barcoded sequences
In the process of multiplex BAC sequencing, DNA barcoding is one of the most laborious steps. It is therefore of substantial interest to quantify the trade-off between experimental effort and the quality of the results. Without barcoding of individual clones, sequencing of a BAC pool, however, results in a single complex assembly of sequences originating from many BACs in contrast to multiple separated assemblies of individual BACs in case of barcoding. This higher complexity is expected to have negative effects on the quality of the non-barcoded assembly due to chimeric contigs derived from different clones based on repetitive elements. To estimate this risk and to answer the question whether barcoded assemblies are superior to non-barcoded ones, we generated three different assemblies of the bcTi sequences from BAC pool 2 (48 non overlapping clones) without separation by barcodes prior assembly ("non-barcoded", non-bc).
Assembly 1 was done with the unmasked reads. For the two other assemblies we used reads which were masked depending on the 20mer frequency of the 454 sequences from BAC pool 2. The reads for assembly 2 ("m72") were masked in regions where the 20mer frequency exceeded 72, corresponding to the 3x mean sequence depth (~24x). Assembly 3 ("m36") was performed with reads masked in regions with a 20mer frequency >36 (1.5x of the mean sequence depth). This resulted in 682, 700 and 761 contigs with a total length of 5.5, 4.5 and 4.0 Mb for the assemblies 1, 2 and 3, respectively.
Statistics of non-chimeric and chimeric contigs >1 kb generated by the assemblies of unmasked and masked bcTi reads of pool 2 without separation by barcodes.
total length (bp)
average length (bp)
fraction of total contig length
To examine chimeric structures in more detail we plotted both the read coverage by different BACs and the 20mer frequency along the chimeric non-bc contigs from the unmasked assembly 1. Visual inspection of these plots revealed that 173 out of the 328 chimeric contigs (53%) consist entirely of repetitive sequences with 20mer frequencies above 100x. The other 155 contigs contain at least one non-repetitive part, showing 20mer frequencies corresponding to the BAC's sequence depth. The non-repetitive contig parts are wrongly joined either to a repetitive or a non-repetitive part from another BAC (additional file 19). The misassembled chimeric regions are characterized by repetitive elements, ranging in length from a few base pairs for tandem repeats up to several kb for long terminal repeats (LTR). For the vast majority of cases, at these points the 20mer frequency considerably exceeds the BAC's sequence depth. We only found one example (contig 35) for which we identified neither increased 20mer frequencies nor known repeat structures at the region of misassembly.
Comparison of NGS data to a high quality Sanger reference is useful to evaluate the trade-off between speed/cost-efficiency and outcome quality. We followed this approach to measure the influence of sequence length and depth on the assembly quality of four barley BACs from barcoded reads and to determine sequencing error rates for the different 454 FLX and Titanium chemistries.
For all reference BACs, the bcTi assemblies were considerably better than the bcFLX ones in terms of consistency and quality. By equalizing sequence depths for both sequencing technologies (bcTi "down sampling"), we could estimate to which extent the read lengths determine these differences. On average, the read lengths of bcFLX and bcTi in our experiments differed by only 40 bp (223 vs 263 bp), but the Titanium chemistry produced long reads with >600 bp (in contrast to only few reads >300 bp generated by FLX). Due to this difference, the Titanium reads create considerably fewer misassemblies (12 vs. 29) and gaps (9 vs. 19) at the same sequence depth compared to FLX. Although this was expected, it has only now been shown for a multiplex approach like the barcoded sequencing of 48-BAC pools. In addition, the effect is surprisingly clear - obviously not due to the relatively modest gain in mean read lengths but rather to the portion of extra-long reads generated by the Titanium platform.
Sequence depth reduction of bcTi from an average of 47x to 23x did not lead to assemblies of lower quality and consistency for three reference BACs,This agrees with our previous observations . With experience sequencing ~3,000 barcoded barley BACs in pools (unpublished data), we can conclude the following: (i) 15x depth is regarded as minimum for an acceptable BAC representation, (ii) depths below ~20x are critical for the assembly quality independent of read length, (iii) coverages much higher than 20x do not improve the assembly quality.
Estimation of sequencing accuracy did not reveal differences between the bcFLX, bcTi and bcTids assemblies (~Q35). About half of the sequence errors are insertions/deletions in homo-nucleotide stretches, illustrating a well known drawback of the pyrosequencing based 454 sequencing method . Another 27% are other insertions or deletions which are mostly embedded or adjacent to homo-nucleotide stretches reflecting the same problem. Single nucleotide changes account for the remaining 27% of sequencing errors. Furthermore, deeper sequence coverages did not improve the overall consensus accuracy, suggesting that a 15 to 20x sequence depth is sufficient in this regard.
The construction of scaffolds consisting of ordered and oriented contigs using MP information is a powerful tool to improve assemblies of previously unordered contigs. We were able to unambiguously arrange 79% of the total contig length of the 96 BACs into 199 scaffolds by Illumina 3 kb MP sequences. This considerably enhanced the assembly quality by more than doubling the L50, L80 and L90 lengths to ~53 kb, ~23 kb and ~8 kb, respectively. By defining a threshold for the minimum number of MPs to reliably bridge gaps, we considered 644 contig pairs. In the resulting graph structures we observed 92 contig ends with more than one edge, for which cases scaffolding was omitted Nevertheless, for 46 branches to two contigs, the normalized numbers of supporting MPs differ by a factor >2. By omitting the low supported branch, 87 additional contigs with a total length of ~1.1 Mb could be scaffolded. This increases the fraction of scaffolded contig length from 79% to 89% (data not shown). This scaffolding rate could presumably be improved further by applying lower or otherwise defined thresholds. Most likely, the small number of additional contigs scaffolded would be paid for by a higher rate of conflicts for which a decision is impossible.
Our scaffolding is based on the mapping of all non-barcoded MPs to each of the 96 barcoded BAC assemblies. This method may result in bridgings of contig pairs from different BACs by the same MP, particularly those from repetitive regions. We checked our data for such doubled occurrence and found that ~5% of the MPs map to contigs from more than one BAC (data not shown). We therefore estimate the risk for wrong scaffolding of assemblies from barcoded BACs by non-barcoded MPs to be low, suggesting that more than 96 BACs can be pooled for the MP libraries. In principle, whole genome shotgun (WGS) derived MPs should also be appropriate to scaffold BAC assemblies. This technique would avoid the preparation and sequencing of customized BAC pools, but on the other hand bear a much higher risk for improper scaffoldings due to repeats. To test this approach we used 3 kb MP sequences from a barley WGS library to scaffold the 454 assemblies from the BAC pools. Onlyy after repeat-masking the MPs could we obtain meaningful but marginal scaffolding (data not shown). We therefore suspect that scaffolding of 454 BAC assemblies by WGS MPs is feasible, but associated with a considerable number of conflicts due to branches. Improved scaffolding may require additional MP distances and a sequence depth substantially higher than in our pilot experiment.
For multiplex sequencing of BACs, particularly those with high repeat contents such as in barley, formation of chimeric contigs represents a major concern. These contigs can be minimized by introducing individual tags prior to sequencing, a process which is laborious and time consuming. To evaluate the impact of barcoding on multiplex BAC sequencing, we assembled sequence data from one 48-BAC-pool and used barcode information to calculate which degree chimeric contigs consist of more than one BAC. With this approach, 47% of the total non-barcoded contig length was identified as chimeric when assembling unmasked sequences.
After assembling repeat masked sequences, the total length of all contigs decreases from ~5.5 Mb for the assembly of unmasked sequences to ~4.0 Mb for the assembly with the highest masking stringency (m36, 20mer frequencies >1.5x of the mean sequence depth). However, repeat masking has nearly no influence on the overall length of non-chimeric contigs (~3 Mb), although the masked assembly is more fragmented (unmasked: 354 contigs/mean 8.2 kb; m36: 562/5.4 kb). In contrast, masking diminishes the number of chimeric contigs (unmasked: 328; m36 199). In the unmasked assembly, more than half of the chimeric contigs consist entirely of repeats (173 out of 328) but only 0.5% (1 out of 199) in the m36 assembly. As a result, the total length of chimeric contigs decreases from ~2.57 Mb in the unmasked to ~0.97 Mb in the m36 assembly, reducing their fraction from 47% to 24%. One can therefore conclude that repeat masking of NGS reads derived from BAC pools prior to de novo assembly reduces the fraction of chimeric contigs by a factor of two. Nevertheless, one quarter of all contigs are still chimeric and would hamper subsequent data interpretation, e.g. gene structure predictions. This observation favours barcoding in NGS of BACs for more consistent assemblies whenever it is feasible.
NGS of BAC pools is a suitable tool for the analysis of large and highly repetitive genomes,. To obtain the most consistent assemblies, large contigs and few gaps, the maximum read length >600 bp of the 454 titanium chemistry is a crucial factor. BAC barcoding is indispensable to assess both repetitive and non-repetitive sequence information due to the high risk of chimeric contig formation during pooled BAC assemblies. When interest is restricted to non-repetitive regions harbouring the majority of genes, repeat masking NGS reads in lieu of barcoding prior to assembly is also an option. In both cases, assemblies can be considerably improved by scaffolding with mate pairs from non-barcoded BAC pools. It remains to be determined whether whole genome mate pair data would also be appropriate for this purpose.
BAC preparation, barcoding, Roche/454 and Illumina/Solexa sequencing
The 4 reference BACs, 43 BACs of pool 1 and all 48 BACs of pool 2 are derived from the same Hordeum vulgare vulgare (cv Morex) library HVVMRXALLhA. 5 BACs of pool1 are derived from different libraries (HVVMRX83KhA, HVVMRXALLe, HVVMRXALLhC, HVVMRXALLrA). For convenience, in the text and tables BAC names are reduced to the last six characters. Full names including the library are listed in additional files 3 and 10.
DNAs were prepared by an adapted "Maxi-Prep" protocol and barcoded after fragmentation as previously described [10, 18]. For FLX sequencing (bcFLX), the reference BACs were part of a 24 barley BAC containing pool which was sequenced by the GS LR70 Sequencing Kit on a half 70 × 75 Picotiterplate on a GS FLX according to the manufacturer's instructions (Roche Diagnostics). For Titanium sequencing (bcTi), the reference BACs were part of a pool of 48 clones, sequenced by the GS Titanium Sequencing Kit XLR70t on a half Titanium 70 × 75 Picotiterplate (additional file 3). Sequencing of the two pools 48 BACs each (additional file 10) by FLX (pool 1) and Titanium (pool 2) chemistries was performed analogously as described for the reference BACs.
An Illumina MP library was constructed of a pool of 96 BACs (pool 3) following the manufacturer's instructions (Illumina). After first fragmentation of the template DNA fragments of ~4,500 bp were excised of the agarose gel. The average fragment length by Agilent DNA 7500 chip was determined to ~4,300 bp. Two lanes of a flow-cell were sequenced on an Illumina GAIIx using Illumina's paired-end cluster generation kit v2 and cycle sequencing kits v4 following the 2 × 36 cycles recipe. Sequences were extracted by the GenomeAnalysis-Pipeline CASAVA v1.6.
Assemblies of 454 sequences and comparison to Sanger references
All assemblies were performed using MIRA version 3.2.0 (http://www.chevreux.org/projects_mira.html) and default parameters with the features "accurate, 454, genome, denovo". In the pre-processing step reads were screened for E. coli and Vector sequences using blastn with a threshold of 10-10. Reads matching the vector sequence >2 kb apart from the restriction site were discarded from the assembly as well as reads with a hit to the E. coli genome. Reads with a vector match up to 2 kb to the restriction site were kept and clipped at the restriction site using a cross match based pipeline. Comparisons of BACs to Sanger references were performed using cross-match (http://www.phrap.org/phredphrapconsed.html) and default parameters. The result was parsed for counting gaps and misassemblies, where a gap was defined as regions in the reference that were not represented in the 454 assembly and a misassembly was defined as two disjunctive parts of the same contig aligning to different regions of the reference.
Scaffolding of 454 contigs using Illumina mate pairs
The distribution of MP distances was determined by mapping to the the contigs from the assembly of BAC 562B07 (pool 2) using bwa (http://bio-bwa.sourceforge.net/bwa.shtml), PE mapping. Minimum and maximum distances were defined as 1.5 inter-quartil-range distance from the quartiles. All Illumina MP reads were separately mapped to the 454-contigs of each BAC by bwa long read mapping. For further analyses only MPs were used of which both reads mapped on different contigs in the right orientation and both in a distance to a contig end according to the maximum distance of 3,742 bp. Pairs of which both reads mapped at exactly the same position at the 454 contigs were regarded as duplicons and reduced to only one pair. Contig pairs that were supported by Illumina MPs were stored in a graph structure using Java Jung library.
Repeats were predicted by k-mer frequencies using Tallymer . The index of frequences was built from all reads of pool2.
Comparison of non-barcoded assembly to barcoded
All reads from pool 2 were assembled without (non-bc) and after separation by barcodes using Mira. Read coverages were extracted in CAF (common assembly format, NCBI) from the non-bc assembly. Graphs were plotted using R. Sequence comparisons were done by the dot-matrix program "dotter" (http://sonnhammer.sbc.su.se/Dotter.html).
List of Abbreviations
Bacterial Artificial Chromosome
BAC end sequences
GS FLX sequencing platform (Roche/454)
Next Generation Sequencing
Titanium sequencing platform (Roche/454)
Whole Genome Shotgun
Acknowledgements and Funding
The work was financially supported by a grant (GABI-BARLEX FKZ0314000) of the German Ministry of Education and Research (BMBF).
We thank Ivonne Heinze, Ivonne Görlich, Kathleen Seitz, Daniela Werler, Ulrike Beier and Anne Kusserow for skillful technical assistance. Bryan Downie is acknowledged for critical proofreading of the manuscript.
- Wicker T, Taudien S, Houben A, Keller B, Graner A, Platzer M, Stein N: A whole-genome snapshot of 454 sequences exposes the composition of the barley genome and provides evidence for parallel evolution of genome size in wheat and barley. Plant J. 2009, 59 (5): 712-722. 10.1111/j.1365-313X.2009.03911.x.PubMedView ArticleGoogle Scholar
- Eversole K, Graner A, Stein N: Wheat and barley genome sequencing. Genetics and genomics of the Triticeae. Edited by: Feuillet C, Muehlbauer J. 2009, Springer, 713-742.View ArticleGoogle Scholar
- Varshney RK, Nayak SN, May GD, Jackson SA: Next-generation sequencing technologies and their implications for crop genetics and breeding. Trends Biotechnol. 2009, 27 (9): 522-530. 10.1016/j.tibtech.2009.05.006.PubMedView ArticleGoogle Scholar
- Arabidopsis_Genome_Initiative: Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature. 2000, 408 (6814): 796-815. 10.1038/35048692.View ArticleGoogle Scholar
- IRGSP: The map-based sequence of the rice genome. Nature. 2005, 436 (7052): 793-800. 10.1038/nature03895.View ArticleGoogle Scholar
- Schulte D, Close TJ, Graner A, Langridge P, Matsumoto T, Muehlbauer G, Sato K, Schulman AH, Waugh R, Wise RP, Stein N: The international barley sequencing consortium--at the threshold of efficient access to the barley genome. Plant Physiol. 2009, 149 (1): 142-147. 10.1104/pp.108.128967.PubMedPubMed CentralView ArticleGoogle Scholar
- Wicker T, Zimmermann W, Perovic D, Paterson AH, Ganal M, Graner A, Stein N: A detailed look at 7 million years of genome evolution in a 439 kb contiguous sequence at the barley Hv-eIF4E locus: recombination, rearrangements and repeats. Plant J. 2005, 41 (2): 184-194.PubMedView ArticleGoogle Scholar
- Wicker T, Schlagenhauf E, Graner A, Close TJ, Keller B, Stein N: 454 sequencing put to the test using the complex genome of barley. BMC Genomics. 2006, 7: 275-10.1186/1471-2164-7-275.PubMedPubMed CentralView ArticleGoogle Scholar
- Rounsley S, Marri PR, Yu Y, He R, Sisneros N, Goicoechea JL, Lee SJ, Angelova A, Kudrna D, Luo M, Affourtit J, Desany B, Knight J, Niazi F, Egholm M, Wing RA: De Novo Next Generation Sequencing of Plant Genomes. Rice. 2009, 2: 35-45. 10.1007/s12284-009-9025-z.View ArticleGoogle Scholar
- Steuernagel B, Taudien S, Gundlach H, Seidel M, Ariyadasa R, Schulte D, Petzold A, Felder M, Graner A, Scholz U, Mayer KF, Platzer M, Stein N: De novo 454 sequencing of barcoded BAC pools for comprehensive gene survey and genome analysis in the complex genome of barley. BMC Genomics. 2009, 10: 547-10.1186/1471-2164-10-547.PubMedPubMed CentralView ArticleGoogle Scholar
- Gonzalez VM, Benjak A, Henaff EM, Mir G, Casacuberta JM, Garcia-Mas J, Puigdomenech P: Sequencing of 6.7 Mb of the melon genome using a BAC pooling strategy. BMC Plant Biol. 2010, 10: 246-10.1186/1471-2229-10-246.PubMedPubMed CentralView ArticleGoogle Scholar
- Feuillet C, Leach JE, Rogers J, Schnable PS, Eversole K: Crop genome sequencing: lessons and rationales. Trends Plant Sci. 2011, 16 (2): 77-88. 10.1016/j.tplants.2010.10.005.PubMedView ArticleGoogle Scholar
- Zonneveld BJ, Leitch IJ, Bennett MD: First nuclear DNA amounts in more than 300 angiosperms. Ann Bot. 2005, 96 (2): 229-244. 10.1093/aob/mci170.PubMedPubMed CentralView ArticleGoogle Scholar
- Goff SA, Ricke D, Lan TH, Presting G, Wang R, Dunn M, Glazebrook J, Sessions A, Oeller P, Varma H, Hadley D, Hutchison D, Martin C, Katagiri F, Lange BM, Moughamer T, Xia Y, Budworth P, Zhong J, Miguel T, Paszkowski U, Zhang S, Colbert M, Sun WL, Chen L, Cooper B, Park S, Wood TC, Mao L, Quail P: A draft sequence of the rice genome (Oryza sativa L. ssp. japonica). Science. 2002, 296 (5565): 92-100. 10.1126/science.1068275.PubMedView ArticleGoogle Scholar
- Palti Y: Rapid and accurate sequencing of the rainbow trout physical map using Illumina technology. Fish Genome Meeting Hinxton, UK. 2011Google Scholar
- Dohm JC, Lottaz C, Borodina T, Himmelbauer H: Substantial biases in ultra-short read data sets from high-throughput DNA sequencing. Nucleic Acids Res. 2008, 36 (16): e105-10.1093/nar/gkn425.PubMedPubMed CentralView ArticleGoogle Scholar
- Ewing B, Green P: Base-calling of automated sequencer traces using phred. II. Error probabilities. Genome Res. 1998, 8 (3): 186-194.PubMedView ArticleGoogle Scholar
- Meyer M, Stenzel U, Hofreiter M: Parallel tagged sequencing on the 454 platform. Nat Protoc. 2008, 3 (2): 267-278. 10.1038/nprot.2007.520.PubMedView ArticleGoogle Scholar
- Kurtz S, Narechania A, Stein JC, Ware D: A new method to compute K-mer frequencies and its application to annotate large repetitive plant genomes. BMC Genomics. 2008, 9: 517-10.1186/1471-2164-9-517.PubMedPubMed CentralView ArticleGoogle Scholar