Keywords

FormalPara List of species

  • Arabidopsis thaliana

  • Brachypodium distachyon

  • Brassica juncea

  • Brassica oleracea

  • Brassica rapa

  • Carica papaya

  • Eucalyptus grandis

  • Fragaria vesca

  • Glycine max

  • Lotus japonicus

  • Malus domestica

  • Medicago truncatula

  • Mimulus guttatus

  • Oryza sativa

  • Populus trichocarpa

  • Prunus persica

  • Solanum tuberosum

  • Sorghum bicolor

  • Theobroma cacao

  • Triticum aestivum

  • Vitis vinifera

  • Zea mays

1 Introduction

Plant genomes are often both highly repetitive with ancient and more recent polyploidy. Because of these features, taxonomic analysis can be a challenge. To comprehensively analyse germplasm resources, allele and haplotype frequencies need to be studied. Next Generation Sequencing (NGS) techniques have greatly advanced genome analysis, enabling the elucidation of the genome or transcriptome sequence of an organism relatively quickly and cheaper in comparison to the traditional Sanger-based sequencing technologies (Edwards et al. 2013). This has opened avenues in plant research, by which genomic variations can be quickly and efficiently determined through computational and statistical analysis.

While NGS data is relatively inexpensive compared to traditional sequence data, sequence quality and read lengths are relatively poor. Accurate estimation of Single Nucleotide Polymorphism (SNP) calling and allele frequencies may result in false positives due to sequence or read mapping error. In some cases, a greater number of biological replicates rather than depth of coverage has been shown to produce better results when calling SNP and allele frequencies (Li 2011). These errors require downstream validation and confirmatory analysis, which usually involves re-sequencing (Kim et al. 2011).

NGS technologies have impacted on plant taxonomic systems, currently leveraging the Linnaean taxonomic methods with DNA barcoding (Pang et al. 2012) and evolutionary studies in plants (Darracq et al. 2010). Methods which make this possible for complex polyploid genomes include targeted sequence capture (Grover et al. 2012) and chromosome flow cytometry (Dolezel et al. 2004). These approaches reduce the complexity of the assembly of NGS data as well as downstream analysis, enabling biologists to answer questions on diversity, polyploid origins, domestication and ancestry of many crops (Zhang et al. 2011).

With the development of advanced sequencing systems, sequence-based SNPs are becoming the marker of choice. The greatest genetic resolution is obtained through the analysis of SNPs and allele frequencies within a population. Two alleles associating with each other by random chance are said to be in Linkage Equilibrium (LE). If this association is found to be non-random either due to physical proximity or selection, they are said to be in Linkage Disequilibrium (LD). LD is the basis of Association Mapping (AM). Accurate AM is highly dependent on the extent of physical LD, population size and population structure (Duran et al. 2010). Small and highly structured populations often lead to elevated false positives in AM studies (Cuesta-Marcos et al. 2010).

The dynamic nature of living organisms, their interaction with the environment, with other species and with each other, affects adaptation and evolution. This led to the “Pangenome” concept (Tetz 2005). This concept proposes that genetic information of all living organisms belongs to a common system, the Pangenome, and that this system is both stable and fluid, and genetic elements such as DNA, RNA and plasmids among others traverse through the system, implying horizontal gene transfer between similar organisms and even across species. Gene transfer has been reported between bacteria (Yue et al. 2012), between viruses and eukaryotes (Wu and Zhang 2011), between plants and prokaryotes, and plants and other eukaryotes (Bock 2010). The concept further postulates that mechanisms of fluidity of the Pangenome include viruses and bacteria through infection, the food chain, death and decay. There has been little advancement of this concept in plants but several studies have been carried out in humans looking at the Asian, Caucasian and African human genomes (Li et al. 2010a) and in bacteria (Hall et al. 2010; Laing et al. 2010). It would be interesting to see applications of this concept to higher plants.

2 Next Generation Sequencing

DNA sequencing technology is undergoing a revolution and at the same time fuelling a revolution in genetics and genomics. Applications for Sanger-based sequencing remain, though the majority of DNA sequencing is now produced by one of a range of NGS technologies. NGS suffers from shorter reads and greater error rates than traditional Sanger sequencing, and its predominance is due to the ability to produce much larger volumes of data at a relatively low cost per sequenced base.

The first commercially available pyrosequencing system was commercialised by Roche (Basel, Switzerland) as the GS20, capable of sequencing over 20 million base pairs (Mbp) in just over 4 h (Margulies et al. 2005). This was replaced in 2007 by the GS FLX model, capable of producing over 100 Mbp of sequence in a similar amount of time. The current system, the GS FLX + can produce around 700 Mbp of data with read lengths of up to 1,000 bp with multiplexing of samples (www.my454.com). The Roche 454 FLX system performs amplification and sequencing in a high-throughput picoliter format. Emulsion PCR enables the amplification of a DNA fragment immobilized on a bead, generating sufficient DNA for the subsequent sequencing reaction. Beads are distributed onto the plate. DNA sequencing involves the sequential flow of both nucleotides and enzymes over the plate, which converts chemicals generated during nucleotide incorporation into a chemiluminescent signal that can be detected by a CCD camera. The light signal is quantified to determine the number of nucleotides incorporated during the extension of the DNA sequence. The output is in the form of ‘flow space’, which is converted to the traditional ACGT nucleotide sequence format. The sequence reads are much longer than most other NGS systems. The main error types are additional or reduced numbers of nucleotides around mononucleotide strings. These errors make the accurate calling of insertion/deletion (indel) differences a challenge.

The Illumina sequencing platforms use reversible terminator chemistry to generate up to 600 Gbp of sequence data per run, the greatest volume of data from any current NGS platform. Sequencing templates are immobilized on a flow-cell surface, and amplification generates clusters of up to 1,000 identical copies of each DNA molecule. Sequencing uses fluorescently labelled nucleotides to produce reads of up to 150 bp in length, though 100 bp is more common. Reads can be produced as pairs. The use of paired reads improves the accuracy of reference mapping, overcoming many of the limitations of short read lengths such as inaccurate resolution of repeats, indels and structural rearrangements. By using the distance between a read pair to infer an insertion or deletion in the reference or sample and to resolve repeats in de novo assembly, higher accuracy is achieved. Illumina sequencing is now becoming the platform of choice for resequencing, SNP discovery, whole-genome shotgun sequencing and de novo assembly (Imelfort et al. 2009b; Imelfort and Edwards 2009; Williams-Carrier et al. 2010; Dong et al. 2011; Shulaev et al. 2011).

The SOLiD System from Life Technologies (Applied Biosystems) enables parallel sequencing of amplified DNA fragments linked to beads. The method uses sequential ligation of dye-labelled oligonucleotides, and the latest 5500xl system produces 20–30 Gbp of data per day, with read lengths of up to 75 bp (www.appliedbiosystems.com/). SOLiD data features a two-base encoding mechanism that interrogates each base twice providing a form of built-in error detection for SNP discovery when comparing reads to a reference.

Ion Torrent is a relatively new technology and uses a high-density array of semiconductor micro reaction chambers (www.iontorrent.com). Changes in pH are recorded as a result of the release of a hydrogen proton during the incorporation of a nucleotide during DNA synthesis. This produces reads of 100-200 bp, with up to 1 Gbp of data per run. The error profile of this system is still unknown, but the technology has potential for cost-effective re-sequencing and variant discovery with fast runs of 2 h.

Pacific Biosciences is one of the first ‘third generation’ sequencing systems to go on the market, and applies a novel single-molecule sequencing technique called SMRT™ (Single Molecule Real Time) technology. Read lengths of around 1,000 bp have been reported (www.pacificbiosciences.com). As with the Ion Torrent system, little is known about the error profile of the system, but missing bases, and hence indel calling would be a likely challenge with this technology.

NGS technologies continue to evolve at an astounding rate and new technologies such as the Oxford Nanopore system are likely to continue to push the market forward over the coming years.

The vast quantities of data generated using NGS require the development of dedicated bioinformatics systems (Edwards et al. 2009; Marshall et al. 2010; Lai et al. 2012c; Lee et al. 2012). It was initially thought that bioinformatics systems would not be able to keep pace with sequencing developments, and while still a bottleneck exists in translating the data into biological information, the growth of bioinformatics has kept track with data production (Batley and Edwards 2009a; Lee et al. 2011b).

3 Sequencing Reference Genomes

There are now several optional approaches to sequence genomes, and the approach undertaken would depend on the use of the genome sequence (Imelfort et al. 2009a; Edwards and Batley 2010). Bacterial Artificial Chromosome (BAC) sequencing is still considered to be the most robust method for genome sequencing and involves the production of an overlapping tiling path of large genomic fragments maintained within BACs. Each BAC is shotgun sequenced, where many short reads are assembled to produce the sequence of the BAC. The whole genome sequence may then be reassembled based on sequence overlaps. This approach, while being the gold standard, remains prohibitively expensive and is unlikely to be undertaken for the majority of germplasm resources.

An alternative approach is the whole-genome shotgun (WGS) method, where the genome is fragmented into millions of smaller reads that are individually sequenced. Computational algorithms then assemble the genome sequence, frequently requiring additional scaffolding or assembly to generate a representative genome. Scaffolding involves the assembly of several overlapping contigs into scaffolds and then assembling these further into a genome assembly. While WGS requires less time and resources than a BAC-by-BAC approach, assembly is often problematic due to repeats within the genome. This is particularly true for many plant species, with polyploid genomes and abundant repetitive elements (e.g. maize, wheat, etc.). With the decreasing cost of sequencing and rapid improvements in both data quality and read length from next and third generation technologies, WGS sequencing projects are likely to be established for many additional germplasm resources.

Arabidopsis thaliana was the first plant species to have a sequenced genome (Arabidopsis_ Genome_ Initiative 2000). Since this milestone, the number of plant genomes being sequenced continues to increase. While several species with small genome sizes as well as model species are now been sequenced, genome researchers are now starting to tackle some of the larger and more complex genomes (Berkman et al. 2011b; Feuillet et al. 2011; Edwards and Wang 2012; Berkman et al. 2013). These can act both as models to understand broad families of plants, as well as reference sequences for the mapping and comparison of unassembled genome sequence data from diverse germplasm resources.

Brassica species share extensive synteny with Arabidopsis thaliana, enabling comparative mapping and exploitation of the Arabidopsis genome sequence for Brassica crop improvement. Among the six cultivated species of Brassica, B. rapa (syn. campestris, AA, n = 10), B. juncea (AABB, n = 18) and B. napus (AACC, n = 19) are agronomically important oilseed crops, whereas B. oleracea (CC, n = 9) provides valuable leafy vegetables (e.g. broccoli, cauliflower, cabbage, kholrabi, etc.). The other two species, B. nigra (BB, n = 8) and B. carinata (BBCC, n = 17) are largely valued as condiments. Proprietary AA, CC and AACC genomes were sequenced in 2009 (http://tinyurl.com/brassicagenome), and recently the multinational Brassica genome project (MBGP) published the first public B. rapa genome (Mun et al. 2010; Wang et al. 2011; Edwards and Wang 2012).

Legumes represent the third largest plant family and are the second most important crop family for the human diet (Cannon et al. 2006). The model species Medicago truncatula is an annual diploid with eight chromosomes. It is closely related to tetraploid alfalfa (M. sativa). A combination of cytogenetic and BAC sequence data show that the M. truncatula genome is organized into distinct gene-rich euchromatin and repeat-rich pericentromeric regions, allowing the M. truncatula genespace to be efficiently sequenced using a BAC-by-BAC strategy (Young et al. 2011). Six chromosomes have been sequenced in a US project and two additional ones were sequenced by partners in Europe. Of special note is the Medicago HapMap Project, which aims to deep-sequence whole genomes of 30 inbred Medicago lines using the Illumina platform, and use the reference genome to determine SNPs and Indels. A current release (Mt3.5) of the genome is available at (www.medicagohapmap.org).

Lotus japonicus is a diploid self-fertile perennial pasture legume, with six chromosomes and a genome sequence of around 450 Mbp. Large-scale genome sequencing of variety Miyakojima MG-20 began in 2000. ESTs, cDNAs and gene segments from Lotus and other legumes were used to select TAC clones, which were then sequenced using a shotgun approach (Sato et al. 2008). Data released is available at www.kazusa.or.jp/lotus. The genome sequences of Medicago truncatula and Lotus japonicus have provided invaluable resources for legume research given that they have both been sequenced using BAC clones making syntenic studies relatively simple and in turn providing evolutionary insights into other species such as Arabidopsis (George et al. 2008; Schlueter et al. 2008; Bertioli et al. 2009).

Glycine max (Soybean) is a major crop that accounts for 70 % of the world’s edible plant-derived protein. Its 1.1 Gbp genome was sequenced using a WGS approach (Schmutz et al. 2010a), and the sequence is available through the phytozome database (http://www.phytozome.net/soybean). Re-sequencing of cultivated and wild varieties has enabled a detailed characterisation of genome variation in this species (Lam et al. 2010). Early and current studies on the evolution and domestication of soybean have shown a loss of genetic diversity as a result of domestication (Hyten et al. 2006; Li et al. 2010b). It has also been hypothesized that there was a single soybean domestication event, although this has not been confirmed since archaeological evidence suggests multiple domestication sites in East Asia (Lee et al. 2011a).

Trees, due to their long life span, have characteristics that distinguish them from annual, herbaceous plants. It is likely that many of these properties are based on a tree-specific genetic foundation. Poplar has been selected as a model system for trees because it has a relatively small genome. Populus trichocarpa (black cottonwood), with a paleopolyploid (2n = 38) genome of approximately 480 Mbp, was selected as the first tree genome to be sequenced. The Populus trichocarpa Nisqually 1 genotype was sequenced using WGS approaches (Tuskan et al. 2006). Approximately 7.6 million reads were assembled into 2,447 scaffolds containing 410 Mb of genomic DNA. The assembly was found to include more than 95 % of known cDNAs. Sequencing this genome allowed for the comparison between perennial and annual plant species on a whole genome basis for the first time and provides resources to help answer tree-specific questions about dormancy, development of a secondary cambium, juvenile-mature phase change and long-term host-pest interactions (Brunner et al. 2004; Tuskan et al. 2004).

Among the hardwoods, the most widely grown is Eucalyptus. Eucalyptus has high economic value due to its fiber, largely exploited for pulp, cellulose and paper. In addition, Eucalyptus has potential as a sustainable biofuel source, due to a fast growth rate. In June 2007 the DOE JGI initiated the Eucalyptus grandis genome-sequencing project. The project started in August 2007, and sequencing was completed by 2009. The project, coordinated by the Eucalyptus Genome Network (EUCAGEN), involved more than 130 scientists from 18 countries. A release of the first annotated assembly is available on phytozome (http://www.phytozome.net/), with an approximate genome size of 691 Mb, 4,952 scaffolds from 32,762 contigs and 300 scaffolds greater than 50 kb in size, representing approximately 94.2 % of the genome.

In 2004, the tomato genome was selected as the reference for sequencing within the Solanaceae genomics project. The sequencing of the domesticated tomato (Solanum lycopersicum) (Consortium 2012) marks the first step in bringing together genetic maps and genomes of all Solanaceae and related plants, including potato, eggplant, pepper, petunia and coffee. The variety ‘Heinz 1706’ was initially selected as BAC resources were already available.,The 900 Mbp genome was sequenced using a combination of sanger and WGS approach. While the genome was known to consist of approximately three-quarters pericentromeric heterochromatin, known to be rich in repetitive sequences and poor in genes, the remaining one-quarter consisted of distal, euchromatic segments of chromosomes that contain mostly single copy sequences and more than 90 % of the genes. As such, only this portion was sequenced (Shibata 2005). To gain further insight into its evolution, its closest wild relative, Solanum pimpinellifolium LA1589 was also sequenced using illumina short reads (Tomato Genome Consortium 2012). The resulting 739 Mb draft genome showed a 0.6 % divergence from the domesticated variety. Comparative studies with potato (Xu et al. 2011c) revealed an 8.7 % nucleotide divergence. Further comparisons with the grape genome supported previous hypothesis that the rosid lineage diverged from a common eudicot ancestor following whole genome triplication. The tomato genome was however shown to have high synteny with potato, pepper, eggplant and nicotiana (Tomato Genome Consortium 2012).

Potato, another member of the family Solanaceae has a variety of ploidy levels, ranging from diploid (2n = 24) to hexaploid (6n = 72), with the cultivated potato varieties being tetraploid. Potato is an economically important food crop with 330 million tons produced globally in 2009 (http://www.fao.org). The size of the genome is 840 Mbp. The potato genome sequencing consortium (PGSC) sequenced the potato genome using a BAC-by-BAC approach (Xu et al. 2011b). Due to its high heterozygosity, the generation of a draft sequence required the use of a homozygous form, called a doubled monoploid (DM), and integration of the sequence with that of a heterozygous diploid form RH89-039-16. The two genomes were used to study the genome structure, with the heterozygous diploid resembling the cultivated tetraploid potato (http://solgenomics.net/organism/Solanum_tuberosum/genome).

The sequencing of grapevine Vitis vinifera (2n = 38) (Jaillon et al. 2007), was performed on the quasi-homozygous genotype PN40024 by a French-Italian collaborative project using a WGS approach. A total of 6.2 million reads representing 8.4x coverage of the genome were assembled with the Arachne12 assembler to produce 316 supercontigs, representing putative allelic haplotypes totaling 11.6 Mbp. The assembly of one of the haplotypes in each heterozygous region resulted in 19,577 contigs and 3,514 supercontigs totaling 487 Mb, consistent with an earlier predicted size of 475 Mb (Lodhi et al. 1995). A different approach to sequencing V. vinifera used a BAC-by-BAC approach (Zharkikh et al. 2008). In a comparison of the two approaches (Zharkikh et al. 2008) showed it was possible to sequence the highly heterozygous genome by using sufficient coverage and a variety of clones with different sizes. This approach yielded more informative data on repetitive elements and useful SNPs.

The monkey flower Mimulus guttatus, has become a model system for studying ecological and evolutionary genetics due to its diverse phenotypes, which include adaptations to desert and aquatic environments, selfing and outcrossing, annual and perennial forms and varied floral morphology. The DOE Joint Genomes Institute (JGI) commenced sequencing Mimulus guttatus in 2006 using a WGS approach. In addition to the WGS sequence, JGI is sequencing 200,000 ESTs each from M. guttatus and M. lewisii. Additionally, WGS sequencing of IM62 inbred lines is ongoing. A draft release of the genome with 321.7 Mb of 2,216 scaffolds, 300.7 Mb of 17,831 contigs with gaps ~ 6.5 % and 512 scaffolds larger than 50 Kb, with 95.7 % of the genome represented in scaffolds greater than 50 Kbp is available on phytozome (http://www.phytozome.net/mimulus).

The papaya-sequencing project was founded by the centre for genomics, proteomics and bioinformatics research Initiative (CGPBRI) at the University of Hawaii in 2004. Papaya (Carica papaya) is the first fruit species and commercially important transgenic plant to be sequenced (Ming et al. 2008). Papaya has nine chromosomes and the size of the genome is 372 Mbp. A WGS approach produced a total of 2.8 million reads generated from a female transgenic cultivar “SunUp”. After filtering, 1.6 million reads were assembled into contigs containing 271 Mb and scaffolds spanning 370 Mb. Validation of the assembly using 16,362 unigenes derived from expressed sequence tags (ESTs), showed 15,064 ESTs (92.1 %) matched the assembly.

Cocoa (Theobroma cacao; 2n = 2x = 20) was sequenced by the international cocoa genome sequencing consortium (ICGS), using a WGS approach with Sanger and Roche 454 technologies (Argout et al. 2011). Assembly was carried out using Newbler software, resulting in 25,912 contigs and 4,792 scaffolds, with an N50 of 473.8 Kb. Illumina reads (x44 coverage) were used to improve the assembly. The total assembly length was 326 Mb, representing approximately 76 % of the estimated genome size of cocoa. The sequencing of cocoa enabled the comparison of the grape, soybean, poplar and A. thaliana genomes with cocoa revealing 682 gene families (2,053 genes) unique to the cocoa genome. This indicated an expansion of some gene families during the evolution of cocoa. Some of the genes were annotated as flavonoid related, a contributing factor to the flavour and scent of chocolate. A re-assembly of the genome with an updated version of the Newbler assembler resulted in an assembly (ICGS Assembly 1.2) with an N50 of 5.624 Mb and the largest scaffold of 18.2 Mb. The new assembly covered 84.3 % of the genome.

Genetic diversity within the Rosaceae required the use of several model species as references for comparative analysis in this family. Model species identified for this purpose include strawberry (Fragaria vesca), peach (Prunus persica) and apple (Shulaev et al. 2008). Due to the complexity of the octoploid cultivated strawberry F. × ananassa, (2n = 8x = 56), the sequencing of its diploid progenitor, the woodland strawberry F. vesca (2n = 2x = 14), was undertaken (Shulaev et al. 2011). A de novo assembly of Roche/454, Life Technologies/SOLiD and Illumina/Solexa platform reads at 39x-combined coverage resulted in over 3,200 scaffolds with an N50 of 1.3 Mb. A total of 95 % (209.8 Mb) of the genome was represented in 272 scaffolds. Resequencing at 26x coverage with Illumina validated the assembly, with 99.8 % of the scaffolds and 99.98 % of bases perfectly matching with Illumina reads.

Sequencing of the apple genome (Velasco et al. 2010) followed a similar approach to that used to sequence the highly heterozygous grape genome Vitis vinifera cv Pinot Noir (Zharkikh et al. 2008) in which a combination of paired end reads produced by Sanger sequencing and unpaired reads produced by sequencing by synthesis was shown to be an efficient way of sequencing and assembling complex heterozygous genomes. In total, 122,146 contigs provided a 16x coverage of the 603 Mb genome, Of these 103,076 were assembled into 1,629 meta-contigs. This assembly consisted of 26 % Sanger paired end reads and 74 % 454 sequencing by synthesis paired and unpaired reads.

Cereal crops diverged from a common ancestor some 60 million years ago and whole-genome organisation exhibits a high degree of synteny (Moore et al. 1995). Rice has the smallest genome size among major cereal crops, estimated at 430 Mbp (Goff et al. 2002a) and the genome sequences of rice provide a basis for integrating and comparing biological information from rice and related cereal crops (Goff et al. 2002b; Yu et al. 2002).

The International Rice Genome Sequencing Project (IRGSP) sequenced an inbred rice cultivar, Oryza sativa ssp. japonica cv. Nipponbare using a clone by clone approach with BACs and PACs. 3,401 BAC and PAC clones were sequenced and assembled, resulting in a high quality reference genome anchored to a genetic map (International Rice Genome Sequencing Project 2005). Gene content analysis estimated 37,544 genes to be present, of which 17,016 were supported by 25,636 full-length cDNAs. Using this reference, analysis of rice agronomic traits was carried out on 50 wild and domesticated rice accessions (Xu et al. 2012). 6.5 million SNPs were identified and population structure analysis determined the domestication origins of rice. Brachypodium is a close relative of the cool season grasses and in 2006 was sequenced by the US Department of Energy Joint Genome Institute (DOE JGI) to provide a genomic bridge between rice and other agronomically important cereals (International Brachypodium Initiative 2010).

The Sorghum bicolor genome consists of approximately 770 Mbp in 10 chromosomes (2n = 20). A WGS approach was applied within the DOE-JGI community sequencing program to sequence this genome. A validation of the assembly by comparison with 27 individually sequenced BACs indicated that the assembly was 98.46 % complete with an error rate of < 1 nucleotide per 10 kb (Paterson et al. 2009). Comparison of the genome sequence with Sorghum ESTs suggests that more than 95 % of known sorghum protein-coding genes are represented in this assembly. Sequencing the sorghum genome opened opportunities for comparative studies to be carried out in the grass family between rice, sorghum, maize and Brachypodium (Gu et al. 2009), including the identification of conserved noncoding sequences (CNSs) between maize, rice and sorghum (Salvi et al. 2007).

Maize was domesticated about 10,000 years ago from the grass teosinte (Doebley et al. 2006). The maize genome consists of about 2.5 Gbp of DNA maintained in 10 chromosomes, which are diverse due to changes in chromatin composition as a result of an increase in long terminal repeat retrotransposons (LTR retrotransposons) (SanMiguel et al. 1998). In 2005 the NSF, the United States Department of Agriculture (USDA), and the United States Department of Energy (DOE) provided 32 million dollars to the Washington University Genome Sequencing Centre, Cold Spring Harbor, the Arizona Genome Institute and Iowa State University, to undertake a maize genome sequencing project. B73 was selected as the maize variety to be sequenced and a BAC-by-BAC approach was chosen to complement the previous maize genome sequencing assessments. The draft release of the maize genome (Schnable et al. 2009) was sequenced using a minimum tiling path of BACs (n = 16,848) and fosmid (n = 63) clones derived from integrated physical, genetic and optical maps. Shotgun sequencing of clones, to 4-6x coverage was completed and sequences manually improved. From this draft sequence, more than 32,000 genes were predicted in the genome, and 99.8 % of these were found to be on reference chromosomes. The majority of the genome space (85 %) was found to contain several hundred transposable element families, spread across the genome.

The size of the wheat (Triticum aestivum) genome is approximately 17,000 Mbp, much larger than related cereal genomes such as barley (Hordeum vulgare, 5,000 Mbp), rye (Secale cereale, 9,100 Mbp) and oat (Avena sativa, 11,000 Mbp). The size and hexaploid nature of the wheat genome create significant problems in elucidating its genome sequence. The International Wheat Genome Sequencing Consortium (IWGSC) (www.wheatgenome.org) was established in 2005 to facilitate and coordinate international efforts toward obtaining the complete sequence of the bread wheat genome. The IWGSC selected the cultivar Chinese Spring as the germplasm source for the project (Gill et al. 2004). A pilot project led by the French National Institute for Agricultural Research (INRA) was initiated in 2004 to assess the BAC fingerprinting of the largest hexaploid wheat chromosome 3B, which has been shown to carry QTLs for disease resistance and wheat quality (Börner et al. 2002; Carter et al. 2012). Using flow cytometry isolated chromosomes (Kubaláková et al. 2002; Doležel et al. 2004), a total of 68,000 BAC clones of a 3B chromosome-specific BAC library (Safar et al. 2004) was fingerprinted at the French National Sequencing Centre, Genoscope, and a minimal tiling path sequenced. This successful isolation and sequencing of chromosome 3B led extensive analysis of homoeologous gene composition and evolution, diversity, recombination and the generation of a physical map for chromosome 3B (Paux et al. 2006; Paux et al. 2008; Horvath et al. 2009; Saintenac et al. 2009; Breen et al. 2010; Hao et al. 2010; Carter et al. 2012). Similarly, chromosome specific BAC libraries have been constructed for chromosomes 1D, 4D and 6D (Janda et al. 2004). To complement these activities, individual flow sorted chromosome arms are being sequenced using Illumina shotgun sequencing (Berkman et al. 2011b; Hernandez et al. 2011; Berkman et al. 2012b; Berkman et al. 2013). Ultimately, under the International Wheat Genome Sequencing Consortium, all 34 wheat chromosome arms will be sequenced (Šafář et al. 2010). While currently these efforts have not produced a finished genome, the assemblies and syntenic builds of individual chromosome arms generated by comparison with related cereals, provides access to genomic sequence for all genes, while placing the majority of genes within an approximate order and orientation. Currently, only data for chromosome 7 is publically available at www.wheatgenome.info (Lai et al. 2012a), but this resource has already provided the basis for chromosome arm specific marker discovery (Nie et al. 2012) (Table 15.1).

Table 15.1 Below summarises the crops sequenced to date. Though not exhaustive, it gives a glimpse into the breath of application of NGS to crop research
Full size table

4 Next Generation Diversity Analysis

While whole genome assemblies provide the most comprehensive resource for understanding an organism, it is currently inconceivable to attempt the de novo assembly of each and every plant species and variant. Following the model of human diversity analysis, after an initial set of diverse individuals was sequenced to provide a reference collection, the focus moved to study genome diversity using whole genome genotyping. With the rapid growth and plummeting cost of sequence data generation, the discovery, association and application of genome diversity information from NGS data is becoming increasingly attractive (Imelfort et al. 2009b; Berkman et al. 2012a). Such diversity studies using NGS data are not without challenges (Duran et al. 2009b). These include the very large data volumes and the high error rates associated with this type of data. However, these challenges are being addressed, and NGS data mining is becoming a common approach for diversity analysis in a range of species (Seeb et al. 2011; Hayward et al. 2012b; Jiang et al. 2012; Kazakoff et al. 2012).

4.1 Reference-Based Diversity Analysis

Diversity analysis using a reference sequence is useful when a well-characterised and annotated genome sequence of a closely related species is available (Duran et al. 2009d). A good reference sequence would ideally be of very high quality, preferably a model organism for a particular genus. Limitations of reference-based diversity analysis include: low level of sequence coverage of the reference genome lowering the resolution and sensitivity with which variations can be identified, assembly and sequencing errors resulting from low-complexity regions and the alignment threshold used for mapping reads to reference. An alternative approach would negate the use of a reference sequence.

4.2 Non Reference-Based Diversity Analysis

Several approaches have been used to access diversity between genomes without the use of a reference. These generally involve sequence comparisons of sequence reads where diversity assessment only takes into account differences between assembled, cultivar-specific reads. This approach has been implemented using transcriptome data in AutoSNP and used to accurately call SNPs (Barker et al. 2003).

5 SNPs

SNPs are the ultimate form of molecular genetic markers, as a nucleotide base is the smallest unit of inheritance. A SNP represents a single nucleotide difference between two individuals at a defined location. There are three different forms of SNPs: transitions (C/T or G/A), transversions (C/G, A/T, C/A, or T/G) or small insertions/deletions (indels) (Edwards et al. 2007a; Hao et al. 2011). SNPs are direct markers as the sequence information provides the exact nature of the allelic variants. Furthermore, this sequence variation can have a major impact on how the organism develops and responds to the environment. SNPs represent the most frequent type of genetic polymorphism and may therefore provide a high density of markers near a locus of interest (Batley and Edwards 2007).

Studies of sequence diversity have recently been performed for a range of plant species. These have indicate that SNPs appear to be abundant in plant systems (Edwards et al. 2007b; Henry and Edwards 2009). SNPs are generally biallelic and only rarely triallelic. This disadvantage, when compared with multiallelic markers is compensated by their relative abundance. The low mutation rate of SNPs makes them excellent markers for the characterisation of germplasm resources (Syvanen 2001). The challenge of SNP discovery is not the identification of polymorphic nucleotide positions, but the differentiation of polymorphisms from abundant sequence errors. This is especially true for NGS data which has a higher error rate than Sanger DNA sequencing. These errors prevent the electronic mining of this data to identify potentially biologically relevant polymorphisms. A major source of sequence error comes from the balance between the need to produce the longest sequence length and the confidence that sequences are called correctly. Because of this, sequence trimming and filtering of sequence data is often performed to reduce the abundance of erroneous sequences (Kircher et al. 2011).

The identification of true polymorphisms in a background of sequence errors can be based on four methods: sequence quality values, redundancy of the polymorphism in an alignment, specificity of an allele call with a variety and co-segregation of SNPs to define a haplotype. By using the various measures of SNP confidence assessment, true SNPs may be identified with reasonable confidence from next generation DNA sequence data.

The frequency of occurrence of a polymorphism at a particular locus provides one of the best measures of confidence in the SNP representing a true polymorphism, and is referred to as the SNP redundancy score (Barker et al. 2003). By examining SNPs that have a redundancy score equal to or greater than two (two or more of the aligned sequences represent the polymorphism), the vast majority of sequencing errors are removed. True SNPs also co-segregate to define a conserved haplotype, however determining haplotypes from short-read data is challenging as sequence reads rarely include multiple SNPs. This is less of an issue for longer sequence reads from the Roche 454 system or in the application of paired reads from the Illumina or ABI SOLiD platforms.

There are many tools available for the discovery of SNPs from NGS data, but few have been designed specifically for plant populations (Appleby et al. 2009; Batley and Edwards 2009b; Duran et al. 2009b; Duran et al. 2013). One tool, based on autoSNP software (Barker et al. 2003; Batley et al. 2003) uses redundancy and haplotype co-segregation for SNP discovery. AutoSNPdb (Duran et al. 2009a) combines the SNP discovery pipeline of autoSNP with a relational database, hosting information on the polymorphisms, cultivars and gene annotations, to enable efficient mining and interrogation of the data. AutoSNPdb was originally developed for Sanger sequence data of rice, barley and Brassica (Duran et al. 2009c), but has also been applied to discover SNPs from wheat 454 data (Lai et al. 2012b) (http://autosnpdb.appliedbioinformatics.com.au/).

In one of the first examples of cereal SNP discovery from next generation genome sequence data, Barbazuk and co-workers identified more than 7,000 candidate SNPs between maize lines B73 and Mo17, with over 85 % validation rate (Barbazuk et al. 2007). This success is particularly impressive considering the complexity of the maize genome and the early version of Roche 454 sequencing applied, which produced an average read length of only 101 bp.

The larger data volumes from the Illumina sequencing platform enable the confident discovery of very large numbers of genome wide SNPs (Imelfort et al. 2009b; Hayward et al. 2012a; Lorenc et al. 2012). More than 1 million SNPs have been identified between six inbred maize lines (Lai et al. 2010). SNPs are more prevalent in diverse germplasm. Around 3.6 million SNPs were identified by sequencing 517 rice landraces (Huang et al. 2010). This study allowed for the association of genome variation with complex traits in rice and is a model for future studies in other species. Allen and coworkers identified 14,078 putative SNPs across representative samples of UK wheat germplasm using Illumina GAIIx sequencing of cDNA libraries (Allen et al. 2011), with a portion of these SNPs validated using KASPar assays (Orrù et al. 2009). Data production for SNP discovery from large genomes remains costly and often requires the development of consortium approaches (Edwards et al. 2012).

5.1 Copy Number Variation

Phenotypic diversity in plants has been attributed to differences resulting from copy number variation (CNV) and SNPs. CNV refers to differences in the number of gene loci between species or cultivars. CNVs can occur at several scales, from single genes to whole genomes. CNVs may be a consequence of previous polyploidy events and are believed to be behind the phenotypic diversity of polyploids, as shown in maize inbred lines (Springer et al. 2009). In addition, CNVs have been shown to originate from gene amplification events (Xu et al. 2011a).

CNVs may account for a greater variation in nucleotide content than SNPs and are believed to be the greatest contributing factor to genetic diversity (Redon et al. 2006; Schnable et al. 2009). Approaches to study CNV and presence-absence variation (PAV) are varied and may involve the use of oligonucleotide microarrays, mRNA or DNA sequences (Springer et al. 2009). Probes from a reference sequence are hybridised with labeled DNA from cultivars of interest and an assessment of their differential hybridisation can be performed to predict copy number. A disadvantage of this approach is that the probes from the reference sequence inherently contain variations, thus confounding analysis. Due to limitations of microarrays, CNV by sequencing (CNV-seq) has been developed (Xie and Tammi 2009). In CNV-seq, DNA fragments of a reference and sample are sequenced and mapped to a template sequence. A sliding window algorithm is then used to count copy numbers per window position. PAV analysis is a relatively new approach to examine diversity between genomes, chromosomes or regions of interest. Unlike CNV, which focuses on quantitative differences of reads between individuals, PAV seeks to identify elements uniquely present or absent in one cultivar irrespective of the gene frequency.

6 Perspectives

As more plant genome sequences become available and the cost of sequencing drops further, attention is shifting to the analysis, interpretation and integration of sequence data through comparative studies. The biggest challenge for highly repetitive genomes remains the resolution of low-complexity regions. Although in the future this may be addressed by emerging sequencing technologies, methods that can effectively address this limitation will greatly advance the study and analysis of large complex polyploid plant genomes. Such methods focus on complexity reduction of repetitive regions. Examples include the sequencing and analysis of low-copy regions of individual wheat chromosome arms (Berkman et al. 2011a) and consensus calling of SNPs based on coverage (Azam et al. 2012), among others. With second generation sequencing technologies becoming cheaper and producing more reads per sequencing run and longer read lengths (Berkman et al. 2012a) coupled with emerging third generation single molecule sequencing technologies which promise even longer read lengths (Lieberman et al. 2010; Rasko et al. 2011) and the increased availability of diverse reference genomes, a broader comparison of germplasm and a greater understanding of plant genome evolution over the coming years will be possible. In addition, the presentation of organism-specific databases with detailed, integrated and intuitive summaries of varied comparative analysis will become more critical and offer plant breeders with highly curated and organised reference resources.