Introduction to databases

Genome databases are essential resources for experimental and computational biologists. Genome-related databases can be broken into two major groups: generalized and specialized (domain) databases. Generalized repositories include the Genbank/DDJB/EMBL databases of nucleic acids sequences and the PIR/SwissProt/PDB protein sequence databases. These resources capture and deliver information on specific classes of molecules, without any phylogenetic or functional restrictions. In contrast, the specialized databases are more limited in scope and are organized around a specific model organism or biological function (tissue type or protein family). Comprehensively, collected sequence data provide essential genomic resources for accelerating molecular understanding of biological properties and for furthering the application of this knowledge.

Often, both types of databases contain a mixture of data from genome projects and supporting studies from the broader scientific community. Although the contributions of the community might lack data consistency and breadth of coverage, these possible deficiencies are offset by the greater expertise behind the individual contributions (from years of focused research). The construction of well-designed bioinformatics platforms and databases allows users to take advantage of diverse data sets and provides a foundation for comparative genomics. Promotion of comparative genomics among model and applied plants allows researchers to grasp the biological properties of each species and to accelerate gene discovery and analyses. Recent progress in plant genomics has discovered and isolated many important genes with functions that increase yield, quality, and tolerance to various environmental stresses.

The TreeGenes database (http://dendrome.ucdavis.edu/TreeGenes) is home to comprehensive genomic data on conifers and other forest tree species (Wegrzyn et al. 2008). It serves as the primary international resource on genetic maps, resequencing, genotyping, and phenotyping in forest trees (Fig. 1). TreeGenes provides custom informatics tools to manage the flood of information resulting from high-throughput genomics projects from sample collection to downstream analysis. This resource is further enhanced with systems that are well connected with federated databases, automated data flows, machine learning analysis, standardized annotations, sequence search tools, and quality control processes. A sample tracking system now sits at the forefront of most large-scale projects. Barcode identifiers assigned to individual trees during sample collection are maintained in the database to identify an individual through DNA extraction, resequencing, genotyping, and phenotyping. Emerging technologies have been applied to integrate a solution for high-throughput SNP discovery in non-model organisms. The Pine Sequence Alignment and SNP Identification Pipeline identifies SNPs from both Sanger and 454 sequencing that reflect true genetic variation (Wegrzyn et al. 2009). The database itself contains ten curated modules that support the storage of data and provide the foundation for web-based searches and visualization tools. DiversiTree (http://dendrome.ucdavis.edu/DiversiTree/), an extensive user-friendly desktop-style interface, queries the TreeGenes database and is designed for bulk data retrieval. It provides the community with access to a multitude of data types including ESTs, primers, trace files, SNPs, individual tree data, genotypes, and phenotypes. DiversiTree also connects directly to the Forest Tree Genetic Stock Center (http://dendrome.ucdavis.edu/ftgsc/) where users can order specific DNA based on sample, sequence, or maker data. The combined resources serve as a powerful knowledge environment for genotype–phenotype information resulting from a multitude of large-scale genomics projects.

Fig. 1
figure 1

This schematic depicts the overall organization of the TreeGenes and tfGDR database systems. The top boxes represent the diverse categories of users that access the system and a few of their research objectives. Despite using different platforms for both the relational database system and the content management software, the sequence data types and resources developed are nearly identical providing opportunity for collaboration in standards and tool development

Full size image

Tree fruit Genome Database Resources (tfGDR, www.tfgdr.org) is a centralized worldwide repository of genomics and genetics data for Rosaceae (www.rosaceae.org) and Citrus (www.citrusgenome.org) species (Jung et al. 2008). TfGDR provides a common bioinformational infrastructure for collecting, integrating and translating the large and diverse amounts of structural, functional, and comparative genomics data into a knowledgebase serving gene discovery, marker-trait identification, and genomics-assisted breeding for tree fruit and related crops (Fig. 1). The database is implemented in Tripal (Ficklin 2011) an open-source framework that blends the power of a web content management system and Generic Model Organism Database (GMOD) chado (www.gmod.org). Tripal is under active development by several collaborating institutions and is part of the GMOD suite of tools. TfGDR serves as a primary resource for access to the genome sequences of apple, peach, mandarin orange, sweet orange, and strawberry and is being further expanded to include the cacao (www.cacaogenomedb.org) and blueberry genome sequences (www.vaccinium.org). Genome sequence, transcriptome, and mapping data are available using the widely used GMOD tools GBrowse (Stein et al. 2002; Donlin 2009), CMap (Youens-Clark et al. 2009), and GBrowse-Syn (Mckay et al. 2010). Custom interfaces provide easy-to-use, clear routes to specific data, and associated data types regardless of database point of entry. Custom computational analysis pipelines are available for functional annotation of genes, transcripts, and markers. Researchers can download, browse or search these datasets using standard interfaces. All sequence datasets in tfGDR are available for searching via standalone NCBI Basic Local Alignment Search Tool (BLAST) and custom batch BLAST tools. Other standalone online tools include sequence assembly and microsatellite marker identification. TfGDR provides a home for community projects through its secure, private Drupal content management interface and aids research collaboration and communication through custom community mailing lists and searchable archives.

In order to provide an effective platform for genomics research, a public database must make data available in a user-friendly, functional manner. When considering the utility of a genome database, four key attributes come to mind: (1) Flexibility allows a system to adapt quickly to changing data types, both in terms of quantity and type. Next generation sequencing, high-throughput genotyping/phenotyping, and large-scale expression studies are common data types. Achieving flexibility involves technical foresight in templates for interfaces and back-end database organization. (2) Standards adhering to community consensus as well as applying formalized nomenclature are important to the organism focus. Describing the function of genes and phenotypes through ontologies, as well as enforcing nomenclatures for genetic markers and novel genes is critical to support comparative analysis. (3) Integration allows for effective data type connections within the database, as well as among related databases. Some of these relationships are derived computationally, while others are manually curated. The strengths of these relationships also vary; some resources share data directly leading to exact mappings between their objects, while others report common attributes of their objects. (4) Interfaces require a great deal of thought and technical input to be both user-friendly and efficient. A repository that is difficult to navigate from a user’s perspective will be quickly abandoned for other resources. Single, form-based searches as well as bulk downloads should be supported for nearly all genomic data types. Examples of how the primary tree fruit and forestry databases are working towards standards and integration between and within their communities are discussed here.

Standards for submission

The field of genomics is faced with the novel file formats, new sequencing technologies, and increasing amounts of data. Genome databases are asked to remain flexible and at the same time adhere to quickly changing community standards. Specialized databases often turn toward the generalized members of the broader genomics community, such as NCBI, for guidance and as the first step for integrating data. Many genetic data forms already have a place for submission within those generalized repositories, including: ESTs, mRNAs, SNPs, polypeptide sequences, and expression data (Table 1). Both TreeGenes and tfGDR regularly import these data types from public repositories and re-package them to provide comparative value to the research community. Some of these enhancements include aligning sequence to existing genomes, re-clustering within species groups, and organizing subsets of sequence data for bulk downloads. For example, TreeGenes and tfGDR provide a custom forest tree and tree fruit annotation pipelines that, in their specificity, improve the characterization of a given sequence. Attempts at collecting this raw sequence data outside or independently of generalized repositories would result in inconsistencies and redundancies downstream. In addition, most reputable publishing groups require the submission of all sequence-based data to the NCBI repository. Reviewers and editors are explicitly directed to check that each data type is submitted in advance of review. This widely accepted requirement has greatly increased the availability of data to genomics researchers. What is lacking in these required sequence submissions is a resource to capture the genotypic, phenotypic, and geographic information that permits much of the analysis in forestry and horticulture genomics. There is currently no resolve or requirement for formal submission of this type of data. This information is generally appended as inconsistently formatted supplemental text files that cannot be easily accessed, directly compared, or readily queried.

Table 1 Repositories for submission of data
Full size table

One role of the specialized database is to develop standards for the data types not included in the larger repositories. The standards for the submission of this information are for the first time aligned with the support of subject journals, such as Tree Genetics and Genomes. The data types in this category include genetic maps, genotype, phenotype, and environmental data. Genetic maps were the first and principal data type for the TreeGenes and tfGDR databases. They serve as the foundation for bridging genomic and phenotypic data. The value of a computational resource of genetic maps is measured by its ability to provide comparison. Packages, such as GMOD’s Comparative Map Viewer (CMap), have provided the foundation for the 62 and 48 genetic maps currently available in the TreeGenes and tfGDR database, respectively. To encourage users to submit their mapping data, a template is provided to record markers and positions along linkage groups according to specific nomenclature requirements. This information is associated with a literature object and is uploaded, reviewed, and assigned a unique accession number. The standardized nomenclature of the markers allows for the generation of automated mappings in the database and comparative map sets within and between tree species. Similar to other sequence databases, the map associated with that accession can be released immediately to the public or held until publication. The accession number is intended for use in manuscripts as a method to uniquely link and identify the map set.

The plans for organization of genotype/phenotype data in TreeGenes/tfGDR developed through two independent USDA/NIFA projects: Marker-Assisted Breeding in Rosaceae (RosBREED) and the Conifer Translation Genomics Network (CTGN). These were very large-scale evaluations of genotypes and phenotypes across multiple species. Some integrated applications are publicly available to bring together phenotypes, genotypes, and resulting associations. NCBI has launched dbGaP (Zhang et al. 2008), a public database to archive genotype and clinical phenotype data from human studies. The Complex Trait Consortium has launched GeneNetwork (Wu et al. 2004), a database for mouse genotype, classical phenotype, and gene expression phenotype data with tools for “per-trait” quantitative trait loci analysis. PhenomicDB is a multi-species solution for non-plant model organisms by Metalife (Groth et al. 2007). None of the existing tools currently handle plant data or geographic information which is vital for the increasing number of landscape genomics studies. Environmental variables such as GPS coordinates, elevation, soil type, and precipitation measures are critical component for many areas of plant comparative genomics. Within TreeGenes, a sample tracking system was developed for CTGN that provides protocols for needle/leaf collection and a mechanism for tracking barcoded tissue samples through the DNA extraction and genotyping process. This same system provides a flexible but well defined interface for individuals to submit location information, environmental descriptors, and phenotypic evaluations (Fig. 2). This interface can gather information from common garden experiments, established breeding plantations and large or small-scale tree samples across a landscape. Recently, the pipeline was expanded to accommodate submissions downstream through unique sample identifiers, tree descriptors, and phenotype values within the range of defined metrics. This information is stored and organized by institution so that studies using the same individual trees can be associated, thereby adding value (and providing for) future analysis. Similar to microarray studies in NCBI’s Gene Expression Omnibus (Edgar et al. 2002), unique accession numbers are assigned to each study and include metadata on the design as well as all related values. Within tfGDR, a prototype breeding data management system has been developed for the WSU Apple Breeding Program. Pedigree, phenotype, and genotype data are uploaded from a template provided by the breeder and stored in the database. Interfaces have been developed to allow the breeder to browse or search this private data by project, site, cultivar, trait, or pedigree and view or download the data in community-agreed standard formats suitable for statistical analysis. Further work will connect this private data to tfGDR public data for the breeder to provide a comprehensive breeders toolbox with cross planning and seedling selection functionality as part of the RosBREED project. This breeding data management system will be expanded, like the TreeGenes sample tracking interface, to gather outside information and provide accessions for fruit tree genotype/phenotype studies.

Fig. 2
figure 2

Development of the sample tracking interface within the TreeGenes database provides a mechanism for submitting barcoded samples and associated each sample with phenotype data collected and genotype data received directly from sequencing centers. Expansion of this system now allows submission of phenotype–genotype data assigned to a unique identifier. This method provides a user the necessary templates to submit their data and obtain an accession number for publication purposes

Full size image

The application of standards does not end with the submission phase. Once data are collected consistently, it must be stored, curated, and annotated in a way that adds value and provides for future extraction and comparative analysis. Ontologies provide a shared and controlled vocabulary that can be used to model the domain in terms of the types of object or concept and their properties and relationships. Ontology is more complex than systematics used for species classification because it involves multiple parents and the opportunity for different relationships. While the structure of an ontology is a strict hierarchy, it is represented by a directed acyclic graph in which multiple types can have parents, with different relationships between them. Ontologies greatly enhance the value of databases by allowing users to query the different databases or subsections of a database using the same keywords and query strings. Gene Ontology (Ashburner et al. 2000) was one of the original tools and was designed around three categories that describe the function(s), process(es), and location(s) of a specific gene’s expression. Plant ontology captures descriptions of organism structures and developmental stages as they relate specifically to plants (Jaiswal et al. 2005). The newer Environmental Ontology (EnvO) describes the habitat of a given organism or population. This descriptive vocabulary links the values of an organism to its environment. Phenotype, Attribute and Trait Ontology (PATO) is an ontology of phenotypic qualities, intended primarily for phenotype annotation and composite phenotypes. PATO is designed to be used in conjunction with ontologies of quality-bearing entities. An example of such an entity is a stem (from plant ontology), which could be the bearer of the quality “dark green” (PATO:0000328). Both TreeGenes and tfGDR maintain and update local copies of these repositories (Fig. 3). Gene Ontology and Plant Ontology have been well integrated into the functional annotation pipelines used by TreeGenes and tfGDR. Much work is yet needed to fully integrate PATO into phenotype submissions and encourage the community to work with these controls to improve the value of phenotypes available. It is the role of the specialized databases to represent the community and submit information back to the ontology consortiums to describe unique plant structures, phenotypes, and phenotype qualities.

Fig. 3
figure 3

Relevant consortium-guided ontologies are stored as local copies within the TreeGenes and tfGDR databases. The vocabularies are applied to genes, plant phenotypes, phenotype qualifiers, and environmental observations as part of the curation and annotation efforts that occur downstream from submission. Application of ontologies improves the ability to extract, exchange, and analyze the sequence data. Contribution of terms back to the ontology consortiums by the fruit tree and forestry databases is important to make species-specific vocabulary available

Full size image

Future directions: toward integration

In preparation for the sequencing of three conifer genomes (Pinus taeda, Pinus lambertiana, and Pseudotsuga menziesii), the primary comparative forestry and fruit tree databases have paired up to apply their vast resources to the challenge of annotating of the largest genomes sequenced. One of the first objectives under this project will be the integration of a web-based annotation tool known as the Genome Sequence Annotation Server (GenSAS) which is being developed to provide fruit tree and Rosaceae researchers access to a collaborative online annotation tool. Through a dedicated secure web interface, users will have access to personalized individual or group analysis space for annotating their genome sequences. This includes the ability to upload and store sequences and analyses, identify and mask repeats, predict open reading frames and microsatellites, run gene prediction programs (or upload output from gene prediction programs), map transcripts from custom EST/cDNA databases and perform protein homology searches against the reference databases SwissProt and TREMBL, and custom protein databases. The results of the analysis are displayed on tracks (Fig. 4) and users then select the evidence they wish to accept on a custom track.

Fig. 4
figure 4

Genome Sequence Annotation Server (GenSAS) is being developed to provide fruit tree and Rosaceae researchers with access to a collaborative online annotation tool. Through a dedicated secure web interface, users will have access to personalized individual or group analysis space for annotating their genome sequences. a Through a dedicated secure web interface, users have access to personalized individual or group analysis space for annotating their genome sequences. This includes the ability to upload and store sequences and analyses, identify and mask repeats, predict open reading frames and microsatellites, run gene prediction programs (or upload output from gene prediction programs), map transcripts from custom EST/cDNA databases and perform protein homology searches against the reference databases SwissProt and TREMBL, and custom protein databases. b The results of the analysis are displayed on tracks and users then select the evidence they wish to accept on a custom track. A core component of GenSAS final development will be full integration with GBrowse to provide a platform for comparative genomics among multiple species

Full size image

This application will be further developed to serve as a tool for the community to curate gene/gene families and provide more comparative analysis across tree and other model plant species. This system will allow researchers to readily curate their gene/gene families of interest from an intuitive interface and take advantage of a comparative platform without an extensive bioinformatics background. A core component of GenSAS final development will be full integration with GBrowse. This web-based genome browser is a product of the GMOD project and has been widely accepted into the genomics community (Stein et al. 2002). The software can readily accept and exchange the GFF3 files that will be used both for the initial manual annotations and the modifications/additions that result from manual curations in GenSAS.

Integration for the purposes of genome annotation is just one area where comparative analysis is necessary. Both TreeGenes and tfGDR use different back-end platforms to house the data that are not directly comparable. In recent years, many databases have overcome this challenge through the application of web services. BioMart is a widely used platform for improving communication between databases sitting on different platforms (Smedley et al. 2009). BioMart enables researchers to perform advanced querying of genomics data sources through a single web interface. The power of the system comes from integrated querying of data sources regardless of their physical locations and database designs. Once these queries have been defined, they may be automated with its “scripting at the click of a button” functionality. These capabilities are extended by integration with various analysis and visualization software packages such as BioConductor, DAS, Galaxy, and Cytoscape. Comparative specialized plant databases, such as Gramene, have also applied this integrated web service functionality to give users access to data housed in other repositories. Both TreeGenes and tfGDR will utilize web services to provide queries to improve the exchange of data and therefore expand the comparative genomics potential of their respective systems.

TreeGenes and tfGDR share many similarities in the type of data, analysis, and informational resources they provide to their respective forest and tree fruit communities. Where they differ provides opportunity for synergistic collaboration. Where tfGDR already has access to several draft or completed tree fruit genome sequences, the two groups can work together to develop a common community-based gene curation system platform that meets the needs of both the forestry and the horticultural community, while eliminating redundancy of effort. Similarly, TreeGenes has more experience in providing users with access to genotype and phenotype data, an approach tfGDR has begun to leverage to their advantage. Together, the two databases can work toward uniform data standards and access that will benefit both research communities and provide greater opportunities in comparative genomics.