- 1 Introduction
- 1.1 Modularity
- 1.2 Ontologies
- 1.3 Associated Software
- 1.4 Complexity and Detail
- 1.5 Data Integration
- 1.6 Support
- 1.7 Modules
- 1.8 Chado Naming Conventions
- 1.9 Design Patterns
- 1.10 View Layers
- 1.11 Inter-schema Bridges
- 1.12 DBMS Functions
The Chado schema has been designed with modularity and compartmentalization of function in mind. Groups of tables concerned with a single knowledge domain are called modules. There is a core module, general, concerned with data underlying all other classes, these tables store information about databases, databases identifiers, and general information about Chado tables. Equal in importance in Chado is cv, the module concerned with controlled vocabularies or ontologies.
All other sets of tables, or modules, link to these general and cv tables directly or indirectly but are limited in scope to specific biological domains. For example, the sequence module is concerned with protein and nucleotide sequence, the pub module is concerned with articles and publications, and so on. In addition to these limitations in scope we see an effective absence of redundancy. For example, there is a module called companalysis, short for ’’computational analysis”. Its tables are responsible for describing algorithms and the output of algorithms. The rad module (for microarrays) uses companalysis in order to refer to algorithms in addition. The uniqueness, and generality, of the modules implies that one can rely on pre-existing modules for function if one is interested in introducing new modules.
Chado should be considered a highly extensible database due to its modular design. The clear segregation of function into modules, or sets of tables, should allow the introduction of new modules covering new domains of knowledge. The fact that new modules have been introduced into Chado since its initial release (e.g. phylogeny, rad, stock) confirms that the authors' design concepts were correct.
One of the more profound, recent changes in the nature of biology has to do with the adoption of ontologies, or controlled vocabularies, as a way to describe and organize data. Our most popular ontologies have arisen from the need to describe the remarkable variety of living things, and are very detailed and broad. Simultaneously these ontologies have served to categorize and classify the contents of entire databases that had been previously been atomized, or only partially coherent. Chado has been built from the outset to integrate with these ontologies, and this feature makes it extremely expressive. The acceptance of ontologies as general standards, and Chado’s use of these ontologies, make Chado an excellent platform for annotation of biological data.
Chado is considered to be one of the key components in the GMOD suite. It should not be considered the database of GMOD, there are other ways of storing data within GMOD (e.g. BioSQL, Bio::DB::GFF, Bio::DB::SeqFeature) but it is the database of choice when complexity and breadth are required. Because of this central position in GMOD it has been the focus of much software development, not just the design of the schema itself but in terms of components using it and adaptor software that connect Chado to some other GMOD component. Thus we are able to use Chado with different browsers such as GBrowse and Apollo (the latter being able to both read from and write to the database). We have a variety of tools that can be used to load sequence data as GFF into Chado and there are different utilities that can move complex data in and out of Chado as XML (XORT, GMODTools).
Complexity and Detail
Part of the impetus for the creation of Chado was the need for a database that could describe all the detail that arises from extensive research done on model organisms. There are a number of schemas available that are built upon on the classic, and simple, concept of the central dogma, from gene to RNA to protein. However it can become difficult to work with simpler schemas when one wants to describe trans-splicing or non-coding genes or other departures from the simple model. The Chado schema is in use at model organism databases such as FlyBase and XenBase and the close ties between the Chado developers and these model organism communities has resulted in a extensively tested data model, capable of handling both detail and departures from the standard model.
One characteristic of biology over the past decades has been the major impact of methods or technologies. With each new adopted approach biologists have created sizable data repositories, either as private collections or in the public domain. The value and meaning of these data types is fully realized when they are tied to other types of data. For example, an evolutionary biologist may be studying the evolution of non-coding, regulatory sequences. In order to do this she will consider integrating expression data from microarrays (rad module) with sequence data from various species (organism module) with phylogenetic trees (phylogeny module) with results from sequence comparison studies (companalysis module). This is not to suggest that such an effort is easy, rather that with a schema like Chado such a proposition is actually possible.
The community using Chado, and GMOD, is extensive and growing. Chado is in use at a number of sites worldwide (see GMOD Users) and is in being considered by researchers in related fields, notably molecular evolution and ecology. The Chado community is active and supports a number of different mailing lists. It is also notable that there are close ties not just between GMOD and Chado users but also between these two groups and the community of ontology developers (for example, OBO, NCBO).
We organised the tables into distinct modular components with tightly defined dependencies. This is recogised as good software engineering practice, it allows different software components to focus on the specific data compartments required. It allows for extensibility and schema evolution within specific modules without disrupting the rest of the schema. Finally, it allows for a mix and match approach - it is the authors' hope that the schema modules will be adopted by other model organism and bioinformatics groups; these groups may want to swap in their own table variants within specific modules, or add modules of their own.
- Audit - for database audits
- Companalysis - for data from computational analysis
- Contact - for people, groups, and organizations
- Controlled Vocabulary (cv) - for controlled vocabularies and ontologies
- Expression - for summaries of RNA and protein expresssion
- General - for identifiers
- Genetic - for genetic data and genotypes
- Library - for descriptions of molecular libraries
- Map - for maps without sequence
- Organism - for taxonomic data
- Phenotype - for phenotypic data
- Phylogeny - for organisms and phylogenetic trees
- Publication (pub) - for publications and references
- Rad - for microarray data
- Sequence - for sequences and sequence features
- Stock - for specimens and biological collections
- WWW -
|sequence||cv general pub organism|
|genetic||sequence cv general pub phenotype|
|expression||sequence cv pub|
|map||sequence cv pub|
|rad||sequence cv pub organism contact general companalysis|
|library||sequence cv pub organism|
|phylogeny||sequence cv pub organism general|
|stock||cv pub general organism genetic|
|www||sequence cv pub phenotype organism expression general genetic|
Inter-module Linking Tables
These can be thought of as floating outside of the respective modules they bridge, although they are generally bundled with one or the other module.
Chado Naming Conventions
We use lowercase in all tables and column names - DBMSs differ in how they treat case sensitivity. For example, Oracle will automatically capitalize everything. So it's best to be neutral and use lowercase.
In table names, we use underscores for linking tables; e.g. feature_dbxref is a linking table between feature and dbxref.
Where a table name is a noun phrase rather than a single noun, we concatenate the words together. For instance the table for describing feature properties is called featureprop. It could be argued this is harder to read, but it does allow consistent usage of underscores as above. FeatureProp could be used where it is known the DBMS is case insensitive.
In column names, we also use concatenated noun phrases, except in the case of primary or foreign keys, e.g. dbxref_id.
We try to keep column names unique where appropriate, which is useful for large join statements or views, in avoiding column name clash between diﬀerent tables. The convention is to use an abbreviated form of the table name plus a noun describing the column, for instance fmin in the feature table. By consistently using abbreviated forms we stop column names getting too big (many DBMSs will complain about long column names).
Primary and foreign key names
We use the same column name for primary and foreign key columns - very useful for NATURAL JOIN statements.
Constraint names are a concatentation of table name, underscore, the letter c, and a digit. For example: feature_phenotype_c1.
Index names are a concatentation of table name, underscore, the string idx, and a digit. For example: feature_phenotype_idx1.
The names of views are lowercase. Where a table name is a noun phrase rather than a single noun, we concatenate the words together using the underscore. For example the view used to query for nucleotide motifs is called nucleotide_motif and the view used to find exons from pseudogenes is called pseudogenic_exon.
Views can be thought of as virtual tables. They provide a powerful abstraction layer over the database. All views should be portable across all DBMSs
Views in chado are deﬁned on a per module basis. View deﬁnitions are maintained in the chado/modules/MODULE-NAME/views directory.
Included in the view directory are report views. These can usually be found in a ﬁle called chado/modules/MODULE-NAME/views/MODULE-NAME-report.sql
Collections of view deﬁnitions are bundled into packages, each package is a .sql ﬁle.
DBMS Functions in Chado are entirely optional.
Functions in chado are deﬁned on a per module basis. Function deﬁnitions are maintained in the chado/modules/MODULE-NAME/functions directory.
Collections of function definitions are bundled into packages. Each package comes with an interface descriptions and one or more implementations.
Function Interface Definitions
The interface descriptions are stored in a *.sqlapi file. The syntax used is a variant of SQL and is intended primarily as a consistent way of providing information for human, although it should be parseable by software.
Here is an example, taken from the top of the chado/modules/sequence/functions/subsequence.sqlapi package. This package provides basic subsequencing functions. It has dependencies on two other function packages, declared at the top of the file. The package declares multiple functions, only the first of which is show here, a function for extracting subsequences from the sequence of a feature.
<sql> IMPORT reverse_complement(TEXT) FROM 'sequtil'; IMPORT get_feature_relationship_type_id(TEXT) FROM 'sequence-cv-helper';
-- basic subsequencing functions --
DECLARE FUNCTION subsequence( srcfeature_id INT REFERENCES feature(feature_id), fmin INT, fmax INT, strandINT )
COMMENT ON FUNCTION subsequence(INT,INT,INT,INT) IS 'extracts a subsequence from a feature referenced by srcfeature_id, within the interbase boundaries determined by fmin and fmax, reverse complementing if strand = -1. The sequence can be DNA or AA. Strand must always by >0 for AA sequences'; </sql>
The goal is to provide implementations for different dialects of procedural SQL. Currently only PostgreSQL dialect is supported. The psql implementations are stored in *.plpgsql files.