Modern plant breeders use genetic marker data to predict phenotypes of novel germplasm, thus increasing the speed and efficiency of breeding while simultaneously reducing cost. These predictions are generated through the use of linked genetic markers to select individuals carrying the favorable alleles (marker-assisted selection [MAS]; e.g., Castro et al., 2003). Alternatively, for more complex traits, models are used to estimate the effects of a large number of markers distributed throughout the genome, which are subsequently used to predict the phenotypes of breeding lines, also known as genomic or genome-wide selection (e.g., Heffner et al., 2009). Both of these methods rely heavily on managing and analyzing large sets of trait and marker data. The rapid advancements in genotyping technology have radically changed the context in which MAS and genomic selection are performed. Therefore, new integrated marker and trait databases are needed to support the complex analyses essential to modern breeding efforts.
Several features of the current genetic marker technologies and the opportunities that they create are driving the demand for integrated marker and trait databases. (i) The size of data sets required for mapping and modeling marker effects is rapidly increasing. Traditionally, biparental mapping studies in barley (Hordeum vulgare L.) involved population sizes of several hundred individuals (e.g., Vales et al., 2005) and maps constructed from around 100 to 200 markers. Recently, approaches to genome-wide association studies (GWAS) used populations of thousands of individuals and over a million markers (e.g., Tian et al., 2011). (ii) Panels of genotypes used for mapping have gone from relatively static biparental mapping populations to ad hoc collections of breeding lines and genotypes that can be manipulated and increased in size by combining multiple data sets—this requires dynamic access to marker and trait data. Multiple users can both generate and analyze these data sets. (iii) Data analysis pathways have become more complex and time dependent to fit within critical breeding steps, resulting in the need for rapid access to appropriate data sets and analysis tools. Methods of analyses for association mapping and genomic selection are evolving and it is critical to have easy access to large data sets to reanalyze data using new approaches. (iv) To connect the results of marker–trait analyses to other plant genomic research it is necessary to link to other plant genomic resources. Taken together, integrated marker and trait databases are a necessary enabling tool to coordinate efforts between breeding programs and to fully capitalize on available genomic tools.
The Hordeum Toolbox (THT) is the barley genotype and phenotype database central to the USDA-National Institute of Food and Agriculture-funded Barley Coordinated Agricultural Project (CAP) (Waugh et al., 2009). Derived from the Germinate database (http://bioinf.scri.ac.uk/public/?page_id=159 [accessed 19 May 2006]; Lee et al., 2005), THT was the first of its kind to integrate state-of-the-art genomics and a multi-institutional collaboration to measure agronomic, morphologic, quality, and disease parameters on large diverse populations as well as of breeder's lines and industry standard varieties over several years. User-defined datasets are delivered as files ready to load onto TASSEL (trait analysis by association, evolution, and linkage) (Bradbury et al., 2007) for association mapping or Flapjack (Milne et al., 2010) for graphical genotyping. Novel tools such as “Cluster by Genotype” and the “Haplotype Viewer” make THT a vital resource for the future of barley genome research. Interconnecting links to plant genomic resources such as PLEXdb (Plant Expression Database) (http://plexdb.org [accessed 19 May 2006]; Wise et al., 2007; Dash et al., 2012), HarvEST (http://harvest.ucr.edu [accessed 19 May 2006]; Close et al., 2007), Gramene (http://gramene.org [accessed 19 May 2006]; Jaiswal et al., 2006) and GrainGenes (http://wheat.pw.usda.gov/GG2/index.shtml [accessed 19 May 2006]; Matthews et al., 2003, Carollo et al., 2005) facilitate access to related contig alignments, oligo probe information, and a variety of gene function annotation data from the NCBI, PlantGDB (Plant Genome DataBase), TAIR (The Arabidopsis Information Resource), or rice (Oryza sativa L.) genome databases.
In this paper we describe the key elements of THT database developed as part of the Barley CAP (Fig. 1), including data curation and upload, main functions, analytical tools developed for THT, and a case study for downloading data. The Hordeum Toolbox is freely available at GitHub (https://github.com/Dave-Matthews/The-Triticeae-Toolbox [accessed 17 Nov. 2010]) and can be adapted to other crop breeding programs.
MATERIALS AND Methods
Database Adaptation: Germinate
The THT database is adapted from the schema for the Germinate database (http://bioinf.scri.ac.uk/public/?page_id=159; Lee et al., 2005), a generic plant data management system implemented in the public domain with the MySQL (My structured query language) relational database and freely available under the terms of the GNU public license (The GNU Operating System, http://www.gnu.org [accessed 19 May 19 2006]). A novel feature of Germinate, unlike any other plant data management systems at the time, was the combination of phenotypic and molecular data derived from germplasm panels, allowing queries among multiple independent data sets containing a common set of lines. Changes were made to the schema in the development of THT to generalize the structure, to deal with synonyms for line accession names and single nucleotide polymorphism (SNP) marker names, and to manipulate experimental metadata.
The Hordeum Toolbox combines curated genotype and phenotype data for lines selected by the 10 participating U.S. breeding programs in the Barley CAP. Each program chose 96 elite breeding lines (F4 or more advanced) each year for 4 yr: in total 3781 lines providing data. These lines were selected to be representative of the germplasm in the breeding program at the time they were submitted. Of the 10 programs, two are winter (Oregon and Virginia) and eight are spring programs; two programs breed exclusively six-row barley (Minnesota and the six-row program in North Dakota) while the others are primarily two-row or a combination (Table 1). Further detail about the participating programs can be found by visiting the THT homepage at http://hordeumtoolbox.org [accessed 16 June 2008], clicking on the “About THT” tab, and selecting “CAP Data Programs.”
|Breeding program||Growth habit||Row type||Lines tested||Agronomic||Malting quality||Disease||Food quality and miscellaneous||Winter adaptation|
|Busch Agricultural Resources, Inc.||Spring||2 and 6||379||87||69||68||3|
|Montana State Univ.||Spring||2||384||7||9||13||3|
|North Dakota State Univ. (2-row)||Spring||2||384||36||19||29||3|
|North Dakota State Univ. (6-row)||Spring||6||384||38||33||29||3|
|Oregon State Univ.||Winter||2 and 6||379||13||18||11||1||3|
|Univ. of Idaho||Spring||2 and 6||382||15||21||13||3|
|Univ. of Minnesota||Spring||6||384||13||21||29||3|
|Utah State Univ.||Spring||2 and 6||383||7||13||12||3|
|Virginia Tech Univ.||Winter||2 and 6||339||22||19||22||3||2|
|Washington State Univ.||Spring||2 and 6||383||24||30||13||3|
All 10 programs planted their 96 chosen entries in replicated yield trials at locations suited to their breeding objectives, and agronomic data (yield, plant height, and heading date) were collected at all sites. Each year, all CAP germplasm lines or subsets of lines (spring, winter, or malting) were evaluated in collaborative trials for 58 other traits (Supplemental File S1). These additional 58 traits fell under the categories of agronomic, malting quality, disease, food quality, and winter growth habit traits (Table 2). In addition, data for other traits were collected when the opportunity arose. These ad hoc traits included incidental occurrence of scald or spot blotch diseases in the field. Trait definitions and measurement units defined by the Plant Ontology (Avraham et al., 2008) were used when possible to promote more robust Triticeae ontologies and broaden the utility of THT.
|Grain yield, kg ha−1||76||70||31||25||5004||2487||1329||1414||All|
|Head drop, 0–9||1||0||0||0||737||0||0||0||Spring|
|Head shattering, %||1||0||0||0||425||0||0||0||Spring|
|Heading date, days||68||67||32||25||5806||3258||1719||1414||All|
|Plant height, cm||72||67||31||25||5249||3314||1329||1414||All|
|Stem length, cm||26||19||16||8||1104||932||710||420||All|
|Straw breakage, %||0||2||0||0||0||55||0||0||Spring|
|Test weight, g L−1||31||22||15||7||2931||997||600||310||All|
|Alpha amylase, 20°DU‡||46||58||16||16||1942||1901||544||580||Malt|
|Barley color, °ASBC§||33||20||16||16||1585||1038||544||580||Malt|
|Beta-glucanase activity, U kg−1 malt||12||6||0||0||491||277||0||0||Malt|
|Beta-glucanase thermostability, U kg−1 malt||12||6||0||0||491||277||0||0||Malt|
|Breeders grain protein, %||0||7||4||1||0||320||205||100||Ad hoc|
|Breeders plump grain, % by weight on 2.4 mm (6/64ʺ) sieve||0||7||4||7||0||320||205||310||Ad hoc|
|Diastatic power, °ASBC||46||58||16||16||1942||1901||544||580||Malt|
|Grain protein, %||55||66||20||16||4000||2107||684||580||Malt|
|Kernel weight, mg||33||20||16||16||1585||1038||544||580||Malt|
|Lipoxygenase activity, U g−1 malt||0||1||1||0||0||287||288||0||Malt|
|Malt beta-glucan, mg kg−1||46||58||16||16||1942||1900||544||580||Malt|
|Malt extract, %||46||58||16||16||1940||1901||544||580||Malt|
|Malt protein, %||0||38||0||0||0||863||0||0||Malt|
|Nondormant seeds, %||1||1||1||0||813||746||718||0||Spring|
|Plump grain, % by weight on 2.4 mm (6/64ʺ) sieve||53||59||22||12||4213||2001||782||440||All|
|Residual beta-glucanase, %||12||6||0||0||492||277||0||0||Malt|
|Soluble protein:total protein, %||46||58||16||16||1940||1901||544||580||Malt|
|Wort color, °ASBC||33||20||16||16||1523||951||531||577||Malt|
|Wort protein, %||46||58||16||16||1940||1901||543||580||Malt|
|Barley yellow dwarf rating, 0–8||1||0||0||0||880||0||0||0||All|
|Common root rot severity, %||1||1||0||0||386||390||0||0||Spring|
|Deoxynivalenol, mg kg−1||3||7||11||8||1161||716||1382||1124||Spring (Midwest)|
|Fusarium head blight (FHB) incidence, %||4||0||0||0||382||0||0||0||Spring|
|FHB reaction type, 0–9||1||2||0||0||96||55||0||0||Spring|
|FHB severity, %||3||6||11||8||1156||1159||1381||1137||Spring (Midwest)|
|Leaf rust, 0–9||6||6||0||0||136||152||0||0||Winter|
|Leaf rust seedling, 0–4, with qualifiers||1||0||0||0||951||0||0||0||All|
|Net blotch, 1–10||11||9||0||0||1072||882||0||0||All|
|Net blotch net form reaction type, 0–9||12||4||0||0||202||110||0||0||Spring|
|Powdery mildew, 0–4||1||0||0||0||913||0||0||0||All|
|Scald, 0–8||3||0||0||0||90||0||0||0||Ad hoc|
|Scald reaction type, 0–9||18||7||0||0||540||133||0||0||Ad hoc|
|Septoria seedling infection response, 0–5||1||2||1||0||958||939||863||0||All|
|Septoria speckled leaf blotch, 0–9||0||2||0||0||0||55||0||0||Ad hoc|
|Spot blotch infection coefficient, %||1||1||1||0||673||771||768||0||Spring|
|Spot blotch reaction type, 0–9||5||0||0||0||23||0||0||0||Ad hoc|
|Spot blotch seedling infection response, 1–9||1||2||1||0||957||939||863||0||All|
|Spot blotch severity, %||4||1||1||0||763||771||768||0||Spring|
|Stripe rust severity, %||4||0||0||0||195||0||0||0||Winter|
|Amylose content, % dry weight basis||0||2||1||0||0||938||864||0||All|
|Grain hardness, SKCS¶||0||2||1||1||0||938||864||859||All|
|Grain width, mm||0||2||1||1||0||938||864||862||All|
|Grain weight, mg||0||2||1||1||0||937||864||862||All|
|Hull proportion, %||0||2||1||1||0||938||803||764||All|
|Kernels per spike||2||0||0||0||769||0||0||0||Spring|
|Phenolic compound content, %||0||2||1||0||0||938||864||0||All|
|Polyphenol oxidase activity, abs#||0||2||1||1||0||938||864||862||All|
|Fall planting heading date, days||2||2||0||0||194||155||0||0||Winter|
|Spring planting heading date, days||2||2||0||0||194||155||0||0||Winter|
|Vernalization score, days||2||2||0||0||194||155||0||0||Winter|
|Winter hardiness, % survival||1||0||0||0||98||0||0||0||Winter|
All breeder's germplasm lines as well as 94 “core” lines that included released cultivars, mapping, and key breeding parents from participating programs (Cuesta-Marcos et al., 2010; Comadran et al., 2011) were genotyped with the Illumina Golden Gate assay (Fan et al., 2003) using two 1536-SNP arrays, barley oligo pool assay (BOPA) 1 and BOPA2 (Close et al., 2009). Data generated in the development of the BOPA SNP arrays from the pilot oligo pool assays (POPAs) (POPA1, POPA2, and POPA3) are also on THT (see Supplemental File S2 for marker sequence, nomenclature, and map position). The POPAs were used to genotype the Steptoe × Morex (Kleinhofs et al., 1993), Morex × Barke (Close et al., 2009), and Oregon Wolfe Barley (OWB) biparental (Costa et al., 2001) doubled haploid mapping populations. These data, along with the BOPA1 panel used to genotype the Haruna Nijo × OHU602 population (Sato et al., 2009), yielded a consensus map containing 2943 SNP loci covering a genetic distance of 1099 cM (Close et al., 2009), which is available at THT along with the Steptoe × Morex, OWB_2383, and Morex × Barke genetic maps. When the allele data were entered into THT, the line names and SNP markers were verified to match entries in the database. Summary data such as the number and percentages of each allele in an experiment are computed by THT and can be used in the user's decision on what data to include in the user-defined dataset.
Barley CAP collaborators conducted their own statistical analysis and reported the means for each genotype and summary statistics for the trials. Germplasm lines are assigned a unique identifier (UID) by THT, as are trial codes, traits, or any other field that will be routinely accessed. The trial code provides a direct link to the experimental annotations and allows variables to be loaded separately from the same trial. For example, agronomic traits could be submitted directly after harvest, with malting quality traits reported at a later time. Comparisons across experiments are facilitated by the use of standard check cultivars and summary statistics for each trial, including the trial mean and number of replications that were measured for each trait. For replicated traits, the standard error of a mean and the probability value for the F test for genotypes from the analysis of variance or mixed model analysis were reported. The range of each variable was determined from the raw data sets. For genotypes that had values that were four or more standard deviations from the mean, results were compared across replications. Potential outliers were identified and inquiries were sent to the collaborator who submitted the data for consideration. The Hordeum Toolbox stores the original data sets (raw data) without modification for archive purposes.
Data Access Levels
The Hordeum Toolbox defines four levels of user access, controlling what data are visible, what data can be added or edited, and what menus are available to manipulate the database. During the data acquisition period of the barley CAP, the public could view most of the data immediately, with the exception of data flagged for a brief delay in general release. Registered CAP participants had permission to view all data when logged onto THT. The curator's access level allowed participants to upload data files and edit existing data on a record-by-record basis, using the data input functions in the Curation Menu. Curation status also allows users to access the Database Menu, which contains tools to generate reports on database content, review the schema, export data, and clean up temporary files. The administrator can grant curator-level access and perform database administration tasks using the Database menu.
Data Curation and Upload
In the development of THT, scripts for data curation were written for direct internet upload. The upload tools provide another layer of checking for trial code tracking, data ranges, line names, line aliases, CAP data programs, marker names, marker synonyms, and trait definitions. The curator can also interact directly with THT to add or modify traits and experiment annotations in real time. Most data load within a few seconds. Errors in the content or structure of the data generate specific messages usually enabling the curator to locate and correct errors in the data quickly. For very large datasets, which can take hours to load, processing is “off-line” with the success or failure of the data load reported via e-mail to the curator.
Analytical Tool Development
The “Cluster by Genotype” analytical tool (Fig. 3) was designed to perform a clustering of a selected set of barley germplasm lines based on their alleles for all markers in the database, resulting in a two-dimensional projection, color coded by cluster. One or more clusters can be selected for further examination such as reclustering or export to TASSEL (Bradbury et al., 2007). For the analysis, Illumina A and B allele calls are replaced by the numerical values 0 and 1, and then missing data points for a marker are imputed as the mean marker score. For clustering, the first two principal component analysis eigenvectors are extracted from the line by marker data matrix. Each barley line is plotted as a point in a scatter graph with eigenvector scores for the x and y coordinates. In parallel, the lines are clustered in two to eight clusters using the partitioning around medoids method (Kaufman and Rousseeuw, 1990). This method was chosen because it is fast and robust. The points in the graph are colored according to the cluster to which the corresponding line belongs. Names of known lines can be given to the analysis to construct a legend allowing clusters to be associated with those lines.
Results and Discussion
After 4 yr of serving as a repository for the Barley CAP project, The Hordeum Toolbox database now contains 4216 germplasm line records, 3781 from the breeding programs and 435 from mapping populations. Of these lines, 4209 have genotype data and 3701 have phenotype data. The Hordeum Toolbox contains sequence information for 5609 molecular markers, 4595 with genotyping data for a total of 14,114,103 genotype data points. Phenotype data from 417 separate experimental trials measuring any of 61 individual traits have provided 159,799 data points (Tables 1 and 2).
The design of THT has enabled researchers to have simple and rapid access to data. Pathways to data by phenotype, breeding program, or year, links to germplasm line and marker selection by user-defined criteria, analytical tools, and quick links to complex searches are all easily accessible from the THT homepage (http://hordeumtoolbox.org; Fig. 4). For simple searches, individual lines, markers, and experiments can be accessed via the quick search text box on the homepage. The strength of THT, however, is the ability to build user-defined sets of data. The four basic data types are germplasm lines, genetic markers, phenotype or genotype experimental trials, and measured traits. The Hordeum Toolbox users can build unique datasets for download containing any or all of the available data for a given category.
Germplasm lines can be selected directly from a pasted list or by interactive menus to select lines by property or by their phenotypic values. Information for each line, available on a summary page, include properties (see below) synonyms, links to the Germplasm Resources Information Network (GRIN) (Wiersema, 1995) when appropriate, pedigree (if available), and links to genotype and phenotype data.
Select Lines by Properties
This interface uses seven categorical variables stored for each line, including the line name or synonym, breeding program, year included in the CAP, and primary end use (malt, food, forage, feed, and/or genetic stock) as well as the genetic properties: growth habit (spring or winter), inflorescence row type (2-row vs. 6-row), and hulled vs. hulless. For this search, any combination of values for any of the variables can be selected. The resulting set of lines can then be selected by the user and is stored in a buffer by THT. Further line selection by properties or phenotypes will add to, replace, or be selected from this original set, depending on the user's selection once additional searches are performed.
Select Lines by Phenotype
This tool allows the user to select sets and subsets of lines based on quantitative traits and allows the selection on a range of values for a trait such as grain protein or heading date. For example, a user can select all lines for all 4 yr with a desirable grain protein value (e.g., 10–11%). They can then search for all lines in all years with a low leaf rust disease incidence (e.g., 0–4). Once those lines are queried, selecting the “Intersect (AND)” radio button on the results page will produce a set of lines fulfilling both parameters.
An individual marker record in THT contains the SNP sequence, nucleotides represented by A or B alleles for BOPA1 and BOPA2 data (Close et al., 2009), and synonyms and annotations to outside databases such as GrainGenes (http://wheat.pw.usda.gov/GG2/index.shtml) and HarvEST (http://harvest.ucr.edu). Lists of markers or their synonyms selected for further analysis can be manually entered into a “Search by Name” text box in the marker selection menu page. Marker sets can also be created by selecting a chromosome from one of the genetic maps in THT, such as the Steptoe × Morex map. When selecting a chromosome, the user can narrow the range of markers to a specific region on the map if desired. Before download, markers can be further narrowed on the basis of minor allele frequency to eliminate those with limited polymorphism and the amount of missing data.
Cluster Lines by Genotype
Evaluation of relatedness or similarity among breeding lines is increasingly done on the basis of genotypic rather than pedigree data. The “Cluster by Genotype” analysis strives to be a visually rich implementation of this evaluation (Fig. 3). Two clear cases for its use are (i) the user has a panel of lines that will serve as a basis for association analysis. To get a feel for the extent of structure in the lines, the user uses Cluster by Genotype to visualize that structure; and (ii) the user has selected lines from different breeding programs knowing that they ultimately want to select a panel of lines from across the different programs that are genetically the most similar, irrespective of their origin. The result is a visually informative presentation of differences and groupings among the currently selected lines. The user may take the next step of choosing one or more clusters and restricting the selected set of lines to the lines in those clusters.
The haplotype viewer allows the THT user to identify barley lines that carry a specific combination of alleles at a limited number of markers, assuming that known major-effect QTL underlie those markers. Although the number of markers is not restricted by THT, the database performance is impacted if many markers are selected at a time. Once the user has selected markers in the THT buffer the “View Haplotypes” button links to a panel of dropdown menus for each marker to specify the desired allele state. At that stage, the user can also select phenotype trials, and the mean of the results for each given line will be displayed next to the allele state (Fig. 2).
User-Defined Datasets for TASSEL and Flapjack
In the last few years, a wealth of genomic data coupled with fast and cheap computational power have enabled plant scientists to begin to look at the entire genome when designing breeding projects and searching for genes of interest. Software packages to assist in this effort are in the public domain and THT has been working with their developers to deliver user-defined data sets to use with these new tools. Specifically, THT can produce downloadable text files that can be directly loaded into TASSEL (Bradbury et al., 2007) for association analysis or loaded into Flapjack (Milne et al., 2010) for graphical genotyping.
The flexibility in designing the files to load into TASSEL (Bradbury et al., 2007) or Flapjack (Milne et al., 2010) reveal the strength and versatility of THT, and the simplicity of the downloaded files enables users of other analysis packages (e.g., JMP Genomics 6.0 (SAS Institute, 2010]) to use the data without extensive manipulation. The ability to create specific subsets of the germplasm lines is particularly important because population structure due to relatedness by descent can bias the statistical association between traits and markers as well as lead to spurious associations (Thornsberry et al., 2001). The gateways to data selection are under the “Quick Links” menu on the THT homepage (see Supplemental File S3 for a step-by-step creation of a TASSEL-ready file and subsequent analysis of results in THT).
Pedigree information is stored as a text string using standard Purdy notation (Purdy et al., 1968). In the first phase of the Barley CAP, pedigrees were manually converted into a tabulated format that could be readily exported and used to calculate co-ancestries among lines. Data input for each cross included the names of both parents, the genetic contributions expected, and the level of inbreeding of each parent. The genetic contribution was assumed to be 0.5 but could be modified to indicate additional generations of backcrossing. To avoid pitfalls, extensive discussions with breeders were necessary, since pedigree records are often inconsistent or incomplete (e.g., intermediate parents in pedigrees are often unknown and levels of inbreeding of those parents were seldom reported). Another complication is that barley varieties used in crosses differ considerably in their level of genetic uniformity. Some may have come from a single spike while others are essentially bulks of partially inbred populations. Although pedigree information is useful, the ability to collect accurate and extensive pedigrees was limiting. Therefore, we chose to discontinue collecting pedigree information and rely on the genetic data.
The THT bioinformatics tool was developed to help plant breeders more easily access and use our crop genetic resources. One of the challenges with this approach has been the tendency for collaborators to use long-established methods for data collection rather than standardized database protocols that would permit analyses across diverse breeding programs. Since its inception, THT has provided data for several barley genomics studies including the analysis of population structure in barley breeding populations using multilocus SNP data (Hamblin et al., 2010) and the assessment of population size and unbalanced data sets on GWAS (Wang et al., 2012). The success and ease of data handling in THT prompted the adoption of the database to the next generation of CAP projects, the Triticeae CAP (T-CAP) (http://www.triticeaecap.org/ [accessed 1 Feb. 2011]), creating The Triticeae Toolbox (T3). The data in the THT database were imported into the barley T-CAP database (T3 Barley) and is available at http://triticeaetoolbox.org/barley (accessed 1 Feb. 2011) while a parallel database for wheat (Triticum aestivum L.) (T3 Wheat) at http://triticeaetoolbox.org/wheat (accessed 1 Feb. 2011) was built on the THT model with slight modifications, primarily in the properties of germplasm lines and measured phenotypic traits.
The flexible structure of THT makes it easy to add new species, traits, and data types into the database. The value of THT and its progeny databases, such as T3 Barley and T3 Wheat, will be continuously enhanced by adding links to rapidly growing statistical tools written in R (R Development Core Team, 2006) and by the addition of enhanced graphical data representation and summarization tools now being written, promising the utility of The Hordeum Toolbox for years to come.
Supplemental Information Available
Supplemental material is available at http://www.crops.org/publications/tpg.
Supplemental File S1. Collaborative trials and large datasets in The Hordeum Toolbox (THT).
Supplemental File S2. Barley oligo pool assay (BOPA) marker nomenclature, map position, and sequence.
Supplemental File S3. Case study with The Hordeum Toolbox (THT) screenshots to guide the user through an exercise to select barley germplasm.