Biological Database Systems

1
Biological Database Systems

• 12.1. Integration of biological data

2
Schema heterogeneity

• Fundamental reason
• Data schemas are developed independently
• Varying structures represent the same or
  overlapping concepts
• Even same or overlapping domains are modeled in
  different ways
• These differences (in schemas describing the same
  domain) are referred to as semantic heterogeneity
• Increasingly important challenge in the broad
  field of data management
• Including specifically scientific (biological)
  data management of course

3
Schema heterogeneity

• Database schemas for the same domain developed by
  independent parties are usually quite different
• Why differing structures?
• People think differently even when faced with the
  same modeling goal
• Example having the same database requirements
  students of database course invariably create
  different database designs (reference here)
• Another example web search interfaces in one
  domain (e.g., car search, hotel search, etc.) are
  very heterogeneous

4
Schema heterogeneity

• Bio-example
• Consider largest nucleotide sequence databases -
  DDBJ (1984), EMBL (1982), GenBank (1983)
• In February, 1986, GenBank and EMBL began a
  collaborative effort (joined by DDBJ in 1987) to
  devise a common feature table format and common
  standards for annotation practice.
• DDBJ/EMBL/GenBank adhere to documented
  guidelines
• The DDBJ/EMBL/GenBank Feature Table Definition
  regulating the content and syntax of the database
  entries
• A set of database policies issued and published
  by the Int. Advisors to DDBJ/EMBL/GenBank
• The International Nucleotide Sequence Database
  Collaboration (INSDC)

5
Schema heterogeneity

• From the practical point of view
• Schema heterogeneity is a problem since it
  requires people who understand meaning of
  combined schemas and people skilled in writing
  transformations (e.g., SQL, XQuery, experts)
• Even more challenging for programs a schema is
  just some text for them they cannot capture
  meaning and intent of the schemas
• Kinds of semantic discrepancies between schemas
• Same schema element in two schemas is given
  different names (e.g., InsertDate and
  SeqInsertDate)
• Attributes in schemas are grouped into table
  structures in different ways
• One schema may cover aspects of the domain not
  covered by the other schema

6
Schema heterogeneity example
7
Schema heterogeneity

• Using standard schemas to resolve semantic
  heterogeneity?
• Create a standard schema describing a particular
  domain and use it
• According to experience standards have limited
  success and only in domains where the incentives
  to agree on standards are quite strong
• But even with standards data providers may share
  their data using a standard and still use their
  original data schemas

8
Data heterogeneity

• Heterogeneity occurs not only in schema
  (semantic-level) but also in data values
  themselves (data-level)
• Multiple ways of referring to the same product,
  gene, protein, etc.
• Multiple ways of describing the same things
  (order and number of address elements)
• Multiple formats (e.g., A,C,T,G,- and IUPAC
  nucleic acid codes alphabets for nucleotide
  sequence data)
• Data incompleteness (e.g., peoples names) and
  uncertainty

9
Integration of bio-databases

• Data integration - providing a unified interface
  to access multiple databases
• At present, more than 1000 bioinformatics
  databases exist
• All types of biological data (e.g., metabolic
  pathways, protein structures, nucleotide
  sequences, diseases, etc.)
• Many biological questions require to
  query/search/integrate data from several sources
• To integrate databases, heterogeneities at the
  various levels have to be overcome
• Technical
• Data-level and semantic-level
• Administrative

10
Integration of bio-databases
From Integration of life science databases by
Kohler, Drug Discovery Today BIOSILICO, 2(2),
2004
11
Integration of bio-databases

• Technical heterogeneity
• Storage
• Flat-file storage
• Relational DBMSs
• Object-oriented DBMSs
• XML data (text2XML convertation)
• Access methods and interfaces
• HTTP/FTP
• Web interfaces
• Web services
• Query languages
• SQL
• OQL (O - object)
• XQuery/XPath

12
Database integration architectures

• Link navigation
• Hyperlink to entries in other databases
• Provide links to the most relevant (known)
  databases

13
Database integration architectures

• Data warehousing
• Integrates and aggregates data of several
  databases into one database
• Best suited for the close integration of a
  limited number of data sources with stable
  database schemas
• XML can solve some problematic issues
• Simplifies schema integration (i.e., each
  database can be have its XML Schema, which
  becomes a part of the large integrated schema)
• Data exchanged in XML-format
• Queries using XQuery/XPath
• BUT, performance/reliability of current XML
  storage systems is not adequate

14
Database integration architectures

• Data warehousing

15
Database integration architectures

• Data warehousing

Metadata
16
Database integration architectures

• Database mediation and federation
• No data is converted to some unique format
• Wrappers (uniform access to heterogeneous data
  sources)
• Integration layer (decomposes user queries, sends
  them to corresponding wrappers, and integrates
  query results from all involved sources)
• Query interface
• Database federations
• (Autonomous) data sources provide their own query
  functionality
• Wrappers mainly translate between different
  interfaces
• Mediated databases
• Wrappers have more active role and implement
  query methods to data sources if necessary

17
Database integration architectures

• Database mediation and federation
• Examples DiscoveryLink from IBM Chap 11 of
  Bioinformatics Managing Scientific Data book,
  K2/BioKleisli Chap 6,8 of Bioinformatics
  Managing Scientific Data book

18
Database integration architectures

• Database mediation and federation

19
Database integration architectures

• Indexing flat-files
• Allows to integrate high number of heterogeneous
  databases
• Principle databases to be integrated are
  provided as flat-files

20
Database integration architectures

• Indexing flat-files
• Integration system uses a script (specific to
  each database) to index flat-files from a
  database
• Script is also responsible for discrimination
  data types and generating links to other relevant
  databases
• Users search several databases via their indexes
• Maintenance of integrated database schema is not
  required
• Adding/removal of any number of databases is easy
• Example Sequence Retrieval System (SRS) Chap 5
  of Bioinformatics Managing Scientific Data book

21
Integration approaches

• Which way of integration is the best?
• Depends on
• Purpose of integration
• E.g., data warehouses may be a good solution for
  compiling all available information on a specific
  topic (e.g., model organism, metabolic pathway,
  etc.) warehouses can even integrate many
  resources if these are semantically closely
  related
• Number of resources
• How often data sources would be added/removed
• How up-to-date data in the integration system
  should be
• Should the functionality (specific tools and
  access methods) of underlying data sources be
  integrated?

22
Integration approaches

• Depends on
• Is mapping of equivalent entries among different
  data sources involved?
• Number of users
• Indexing flat-files system can handle many
  simultaneous users while federated databases can
  certain limits on a number of users
• Complexity of schemas to be integrated
• Restriction applied to the usage of the data
• Legal/protective issues some sources cannot be
  fully distributed (i.e., it might be illegal to
  retreive all the data from a particular resource
  and use it in a warehouse system)
• E.g., it might be required to always display the
  origin of the data source

23

• References
• Integration of life science databases by Kohler,
  Drug Discovery Today BIOSILICO, 2(2), 2004
• Bioinformatics Managing Scientific Data by
  Lacroix Critchlow, Morgan Kaufmann, 2003
  (ISBN-10 155860829X)