Abstract

The decreasing cost along with rapid progress in next-generation sequencing and related bioinformatics computing resources has facilitated large-scale discovery of SNPs in various model and nonmodel plant species. Large numbers and genome-wide availability of SNPs make them the marker of choice in partially or completely sequenced genomes. Although excellent reviews have been published on next-generation sequencing, its associated bioinformatics challenges, and the applications of SNPs in genetic studies, a comprehensive review connecting these three intertwined research areas is needed. This paper touches upon various aspects of SNP discovery, highlighting key points in availability and selection of appropriate sequencing platforms, bioinformatics pipelines, SNP filtering criteria, and applications of SNPs in genetic analyses. The use of next-generation sequencing methodologies in many non-model crops leading to discovery and implementation of SNPs in various genetic studies is discussed. Development and improvement of bioinformatics software that are open source and freely available have accelerated the SNP discovery while reducing the associated cost. Key considerations for SNP filtering and associated pipelines are discussed in specific topics. A list of commonly used software and their sources is compiled for easy access and reference.

Go to:

1. Introduction

Molecular markers are widely used in plant genetic research and breeding. Single Nucleotide Polymorphisms (SNPs) are currently the marker of choice due to their large numbers in virtually all populations of individuals. The applications of SNP markers have clearly been demonstrated in human genomics where complete sequencing of the human genome led to the discovery of several million SNPs [1] and technologies to analyze large sets of SNPs (up to 1 million) have been developed. SNPs have been applied in areas as diverse as human forensics [2] and diagnostics [3], aquaculture [4], marker assisted-breeding of dairy cattle [5], crop improvement [6], conservation [7], and resource management in fisheries [8]. Functional genomic studies have capitalized upon SNPs located within regulatory genes, transcripts, and Expressed Sequence Tags (ESTs) [9, 10]. Until recently large scale SNP discovery in plants was limited to maize, Arabidopsis, and rice [11–15]. Genetic applications such as linkage mapping, population structure, association studies, map-based cloning, marker-assisted plant breeding, and functional genomics continue to be enabled by access to large collections of SNPs. Arabidopsis thaliana was the first plant genome sequenced [16] followed soon after by rice [17, 18]. In the year 2011 alone, the number of plant genomes sequenced doubled as compared to the number sequenced in the previous decade, resulting in currently, 31 and counting, publicly released sequenced plant genomes (http://www.phytozome.net/). With the ever increasing throughput of next-generation sequencing (NGS), de novo and reference-based SNP discovery and application are now feasible for numerous plant species.

Sequencing refers to the identification of the nucleotides in a polymer of nucleic acids, whether DNA or RNA. Since its inception in 1977, sequencing has brought about the field of genomics and increased our understanding of the organization and composition of plant genomes. Tremendous improvements in sequencing have led to the generation of large amounts of DNA information in a very short period of time [19]. The analyses of large volumes of data generated through various NGS platforms require powerful computers and complex algorithms and have led to a recent expansion of the bioinformatics field of research. This book chapter focuses on the a priori discovery of SNPs through NGS, bioinformatics tools and resources, and the various downstream applications of SNPs.

Go to:

2. History and Evolution of Sequencing Technologies

2.1. Invention of Sequencing

In 1977, two sequencing methods were developed and published. The Sanger method is a sequencing-by-synthesis (SBS) method that relies on a combination of deoxy- and dideoxy-labeled chain terminator nucleotides [20]. The first complete genome sequencing, that of bacteriophage phi X174, was achieved that same year using this pioneering method [21]. The chemical modification followed by cleavage at specific sites method also published in 1977 [22] quickly became the less favored of the two methods because of its technical complexities, use of hazardous chemicals, and inherent difficulty in scale-up. In contrast, the Sanger method, for which Frederick Sanger was awarded his second Nobel Prize in chemistry in 1980, was quickly adopted by the biotechnology industry which implemented it using a broad array of chemistries and detection methods [19].

2.2. Sequencing Technologies

In the last decade, new sequencing technologies have outperformed Sanger-based sequencing in throughput and overall cost, if not quite in sequence length and error rate [23]. This section will focus on the three main NGS platforms as well as the two main third-generation sequencing (TGS) platforms, their throughput and relative cost. We made every effort to ensure the accuracy of the data at the time of submission. However, the cost and throughput of these sequencing platforms change rapidly and, as such, our analysis only represents a snapshot in time. The flux of innovation in this field imposes a need for constant assessment of the technologies' potentials and realignment of research goals.

2.2.1. Roche (454) Sequencing

Pyrosequencing was the first of the new highly parallel sequencing technologies to reach the market [24]. It is commonly referred to as 454 sequencing after the name of the company that first commercialized it. It is an SBS method where single fragments of DNA are hybridized to a capture bead array and the beads are emulsified with regents necessary to PCR amplifying the individually bound template. Each bead in the emulsion acts as an independent PCR where millions of copies of the original template are produced and bound to the capture beads which then serve as the templates for the subsequent sequencing reaction. The individual beads are deposited into a picotiter plate along with DNA polymerase, primers, and the enzymes necessary to create fluorescence through the consumption of inorganic phosphate produced during sequencing. The instrument washes the picotiter plate with each of the DNA bases in turn. As template-specific incorporation of a base by DNA polymerase occurs, a pyrophosphate (PPi) is produced. This pyrophosphate is detected by an enzymatic luminometric inorganic pyrophosphate detection assay (ELIDA) through the generation of a light signal following the conversion of PPi into ATP [25]. Thus, the wells in which the current nucleotides are being incorporated by the sequencing reaction occurring on the bead emit a light signal proportional to the number of nucleotides incorporated, whereas wells in which the nucleotides are not being incorporated do not. The instrument repeats the sequential nucleotide wash cycle hundreds of times to lengthen the sequences. The 454 GS FLX Titanium XL+ platform currently generates up to 700 MB of raw 750 bp reads in a 23 hour run. The technology has difficulty quantifying homopolymers resulting in insertions/deletions and has an overall error rate of approximately 1%. Reagent costs are approximately $6,200 per run [26].

2.2.2. Illumina Sequencing

Illumina technology, acquired by Illumina from Solexa, followed the release of 454 sequencing. With this sequencing approach, fragments of DNA are hybridized to a solid substrate called a flow cell. In a process called bridge amplification, the bound DNA template fragments are amplified in an isothermal reaction where copies of the template are created in close proximity to the original. This results in clusters of DNA fragments on the flow cell creating a “lawn” of bound single strand DNA molecules. The molecules are sequenced by flooding the flow cell with a new class of cleavable fluorescent nucleotides and the reagents necessary for DNA polymerization [27]. A complementary strand of each template is synthesized one base at a time using fluorescently labeled nucleotides. The fluorescent molecule is excited by a laser and emits light, the colour of which is different for each of the four bases. The fluorescent label is then cleaved off and a new round of polymerization occurs. Unlike 454 sequencing, all four bases are present for the polymerization step and only a single molecule is incorporated per cycle. The flagship HiSeq2500 sequencing instrument from Illumina can generate up to 600 GB per run with a read length of 100 nt and 0.1% error rate. The Illumina technique can generate sequence from opposite ends of a DNA fragment, so called paired-end (PE) reads. Reagent costs are approximately $23,500 per run [26].

2.2.3. Applied Biosystems (SOLiD) Sequencing

The SOLiD system was jointly developed by the Harvard Medical School and the Howard Hughes Medical Institute [28]. The library preparation in SOLiD is very similar to Roche/454 in which clonal bead populations are prepared in microreactors containing DNA template, beads, primers, and PCR components. Beads that contain PCR products amplified by emulsion PCR are enriched by a proprietary process. The DNA templates on the beads are modified at their 3′ end to allow attachment to glass slides. A primer is annealed to an adapter on the DNA template and a mixture of fluorescently tagged oligonucleotides is pumped into the flow cell. When the oligonucleotide matches the template sequence, it is ligated onto the primer and the unincorporated nucleotides are washed away. A charged couple device (CCD) camera captures the different colours attached to the primer. Each fluorescence wavelength corresponds to a particular dinucleotide combination. After image capture, the fluorescent tag is removed and new set of oligonucleotides are injected into the flow cell to begin the next round of DNA ligation [19]. This sequencing-by-ligation method in SOLiD-5500x1 platform generates up to 1,410 million PE reads of 75 + 35 nt each with an error rate of 0.01% and reagent cost of approximately $10,500 per run [26].

Although widely accepted and used, the NGS platforms suffer from amplification biases introduced by PCR and dephasing due to varying extension of templates. The TGS technologies use single molecule sequencing which eliminates the need for prior amplification of DNA thus overcoming the limitations imposed by NGS. The advantages offered by TGS technology are (i) lower cost, (ii) high throughput, (iii) faster turnaround, and (iv) longer reads [19, 29]. The TGS can broadly be classified into three different categories: (i) SBS where individual nucleotides are observed as they incorporate (Pacific Biosciences single molecule real time (SMART), Heliscope true single molecule sequencing (tSMS), and Life Technologies/Starlight and Ion Torrent), (ii) nanopore sequencing where single nucleotides are detected as they pass through a nanopore (Oxford/Nanopore), and (iii) direct imaging of individual molecules (IBM).

2.2.4. Helicos Biosciences Corporation (Heliscope) Sequencing

Heliscope sequencing involves DNA library preparation and DNA shearing followed by addition of a poly-A tail to the sheared DNA fragments. These poly-A tailed DNA fragments are attached to flow cells through poly-T anchors. The sequencing proceeds by DNA extension with one out of 4 fluorescent tagged nucleotides incorporated followed by detection by the Heliscope sequencer. The fluorescent tag on the incorporated nucleotide is then chemically cleaved to allow subsequent elongation of DNA [30]. Heliscope sequencers can generate up to 28 GB of sequence data per run (50 channels) with maximum read length of 55 bp at ~99% accuracy [31]. The cost per run per channel is approximately $360.

2.2.5. Pacific Biosciences SMART Sequencing

The Pacific Biosciences sequencer uses glass anchored DNA polymerases which are housed at the bottom of a zero-mode waveguide (ZMW). DNA fragments are added into the ZMW chamber with the anchored DNA polymerase and nucleotides, each labeled with a different colour fluorophore, and are diffused from above the ZMW. As the nucleotides circulate through the ZMW, only the incorporated nucleotides remain at the bottom of the ZMW while unincorporated nucleotides diffuse back above the ZMW. A laser placed below the ZMW excites only the fluorophores of the incorporated nucleotides as the ZMW entraps the light and does not allow it to reach the unincorporated nucleotides above [32]. The Pacific Biosciences sequencers can generate up to 140 MB of sequences per run (per smart cell) with reads of 2.5 Kbp at ~85% accuracy. The cost per run per smart cell is approximately $600.

Among the TGS technologies, Pacific Biosciences SMART and Heliscope tSMS have been used in characterizing bacterial genomes and in human-disease-related studies [31]; however, TGS has yet to be capitalized upon in plant genomes. The Heliscope generates short reads (55 bp) which may cause ambiguous read mapping due to the presence of paralogous sequences and repetitive elements in plant genomes. The Pacific Biosciences reads have high error rates which limit their direct use in SNP discovery. However, their long reads offer a definite advantage to fill gaps in genomic sequences and, at least in bacterial genomes, NGS reads have proven capable of “correcting” the base call errors of this TGS technology [33–36]. Hybrid assemblies incorporating short (Illumina, SOLiD), medium (454/Roche), and long reads (Pac-Bio) have the potential to yield better quality reference genomes and, as such, would provide an improved tool for SNP discovery.

The choice of a sequencing strategy must take into account the research goals, ability to store and analyze data, the ongoing changes in performance parameters, and the cost of NGS/TGS platforms. Some key considerations include cost per raw base, cost per consensus base, raw and consensus accuracy of bases, read length, cost per read, and availability of PE or single end reads. The pre- and postprocessing protocols such as library construction [37] and pipeline development and implementation for data analysis [38] are also important.

2.3. RNA and ChIP Sequencing

Genome-wide analyses of RNA sequences and their qualitative and quantitative measurements provide insights into the complex nature of regulatory networks. RNA sequencing has been performed on a number of plant species including Arabidopsis [39], soybean [40], rice [41], and maize [42] for transcript profiling and detection of splice variants. RNA sequencing has been used in de novo assemblies followed by SNP discovery performed in nonmodel plants such as Eucalyptus grandis [43], Brassica napus [44], and Medicago sativa [45].

RNA deep-sequencing technologies such as digital gene expression [46] and Illumina RNASeq [47] are both qualitative and quantitative in nature and permit the identification of rare transcripts and splice variants [48]. RNA sequencing may be performed following its conversion into cDNA that can then be sequenced as such. This method is, however, prone to error due to (i) the inefficient nature of reverse transcriptases (RTs) [49], (ii) DNA-dependent DNA polymerase activity of RT causing spurious second strand DNA [50], and (iii) artifactual cDNA synthesis due to template switching [51]. Direct RNA sequencing (DRS) developed by Helicos Biosciences Corporation is a high throughput and cost-effective method which eliminates the need for cDNA synthesis and ligation/amplification leading to improved accuracy [52].

Chromatin immunoprecipitation (ChIP) is a specialized sequencing method that was specifically designed to identify DNA sequences involved in in vivo protein DNA interaction [53]. ChIP-sequencing (ChIP-Seq) is used to map the binding sites of transcription factors and other DNA binding sites for proteins such as histones. As such, ChIP-Seq does not aid SNP discovery, but the availability of SNP data along with ChIP-Seq allows the study of allele-specific states of chromatin organization. Deep sequence coverage leading to dense SNP maps permits the identification of transcription factor binding sites and histone-mediated epigenetic modifications [54]. ChIP-Seq can be performed on serial analysis of gene expression (SAGE) tags or PE using Sanger, 454, and Illumina platforms [55, 56].

The DNA, RNA, and ChIP-Seq data is analysed using a reference sequence if available or, in the absence of such reference, it requires de novo assembly, all of which is performed using specialized software, algorithms, pipelines, and hardware.

Go to:

3. Computing Resources for Sequence Assembly

The next-generation platforms generate a considerable amount of data and the impact of this with respect to data storage and processing time can be overlooked when designing an experiment. Bioinformatics research is constantly developing new software and algorithms, data storage approaches, and even new computer architectures to better meet the computation requirements for projects incorporating NGS. This chapter describes the state-of-the-art with respect to software for NGS alignment and analysis at the time of writing.

3.1. Software for Sequence Analysis

Both commercial and noncommercial sequence analysis software are available for Windows, Macintosh, and Linux operating systems. NGS companies offer proprietary software such as consensus assessment of sequence and variation (Cassava) for Illumina data and Newbler for 454 data. Such software tend to be optimized for their respective platform but have limited cross applicability to the others. Web-based portals such as Galaxy [57] are tailored to a multitude of analyses, but the requirement to transfer multigigabyte sequence files across the internet can limit its usability to smaller datasets. Commercially available software such as CLC-Bio (http://www.clcbio.com/) and SeqMan NGen (http://www.dnastar.com/t-sub-products-genomics-seqman-ngen.aspx) provide a friendly user interface, are compatible with different operating systems, require minimal computing knowledge, and are capable of performing multiple downstream analyses. However, they tend to be relatively expensive, have narrow customizability, and require locally available high computing power. A recent review by Wang et al. [58] recommends Linux-based programs because they are often free, not specific to any sequencing platform, and less computing power hungry and, as a consequence, tend to perform faster. Flexibility in the parameter's choice for read assembly is another major advantage. However, most biologists are unfamiliar with Linux operating systems, its structure and command lines, thereby imposing a steep learning curve for adoption. Linux-based software such as Bowtie [59], BWA [60], and SOAP2/3 [61] have been used widely for the analysis of NGS data. Other software may not have gained broad acceptance but may have unique features worth noting. For reviews on NGS software, see Li and Homer [62], Wang et al. [58], and Treangen and Salzberg [63]. Characteristics of the most common NGS software and their attributes are listed in Table 1, and their download information can be found in Table 4.