1. The LIBI Federated database
“GRID Computing” Tutorial
17 September 2008

2. Agenda Bioinformatic Data integration issues in grid environments
Data Federation in the LIBI platform
Tutorial goals:
a simple case study for querying the federated DB
Designing a data abstraction model on bioinformatic information: the DDQB application

3. Issues Concerning Data Integration
In the bioinformatics’ domain, an increasing number of grid applications manage data at very large scales of both size and distribution.
The complexity of data management on a grid arises from the scale, dynamism, autonomy, heterogeneity and distribution of data sources.
Mission to accomplish:
The goal is providing an IT layer that allows grid applications to access data without taking into account issues such as those exposed before (large scale distribution, dynamism, heterogeneity, etc.)
Viable approaches:
Data Federation: data are logically integrated
Data Warehouse: data are physically integrated

4. Data Federation vs Data WarehouseDF DW
Data warehousing ‘cleaning up’ data and placing it into a centralized repository works well in situations where data are relatively static and data types are not too different
Moving data into a warehouse can limit the specialized search capabilities available with (through) the original data source.
Building and maintaining enterprise wide warehouse on the scale required by most large research organizations with hundreds of data sources can be both costly and risky to implement.
Data warehouse centralization clash with the basic grid-concepts of data replication and distribution according to monitored statistics
Data federation allow to access current data from multiple, heterogeneous, dislocated data sources simultaneously, with a single query
For bioinformatic problems Data Federation seems the most promising solution

5. Data Federation Layer LIBI Federated Database
Name
LIBI Federated Database
Institution
IBM Innovation Lab – Bari
Service
Content
Relational DB federating MitoRes, UTRef, UTRSite, Pubmed, GenBank, OMIM, Uniprot, HmtDB, EMBL_CDS
Description
This database federates local and remote resources to provide a uniform, standard interface to access data.
This federated DB consist of a rationalized composition of both ITB-owned DBs (MitoRes, UTRef, UTRSite) stored locally, of the resource HmtDB owned by Biology Dept. of Bari University, and public NCBI DBs accessed remotely (Pubmed, GenBank, OMIM). The Uniprot DB replicated locally has been federated as well.
Federation provides unique inter-database relationship features that enable users to discover and extract relevant pieces of information with a single query, even if they would be originated from different DBs.
Besides this prominent advantage for the LIBI-platform and its end-users, federated DB provides developers and service consumers with a homogeneous, standard interface (the SQL language) to access data stored locally and/or remotely.
Availability
For LIBI users and services
Access
DB2 DRDA service available at
Main components
IBM DB2, WebSphere Federator Server (with relational & non-relational wrappers)
Custom components
EMBL/FASTA wrapper
Average user access
Tens of concurrent users
DBMS
IBM DB2 9.1
Query language
SQL
Local DBs
MitoRes, UTRef, UTRSite, Uniprot, EMBL_CDS
Remote DBs
Pubmed, GenBank, OMIM, HmtDB

6. Accessing Federated Data: WebSphere Federator Server and related activities
WebSphere Federator Server enables federated databases to access heterogeneous and distributed (local and remote) data sources.
It provides a unified data-management interface (for query and insert): the SQL language
Other features of the federated DB:
DB user-defined functions have been built to accomplish bioinformatics-specific data handling and analyses
A web graphical interface to access federated DB information has been made available (DDQB)
We are developing new interfaces to expose Federated DB capabilities as Web Services
Collaborations:
SPACI developed a wrapper to the DB2 for GRelC, to access the federated DB from Grid environments

7. Data sources integrated in the Federated LIBI Schema
DBName
Data source
location
connection/type
DBName
Data source
location
connection/type
GenBank
Web Services
Species 2000
Web Services
PubMed
Web Services
Mitores
MySQL
OMIM
Web Services
UTRef
MySQL
HmtDB
DB2
UTRSite
MySQL
Uniprot
LEGENDA
relational database
flat file
web services

8. Overall Federated DB Schema
UTRef
MitoRes
UTRSite
GenBank
OMIM
PUBMED
HmtDB
Uniprot
EMBL_CDS

9. TUTORIAL: a simple case study involving federated DB (1/4)
Need
Studying the regulation of the expression of the topoisomerase I at mRNA level.
Biologist needs to retrieve information about regulation of expression of the topoisomerase I. To this end she decides to investigate if there are UTR sequences responsible for the regulation of the expression of the corresponding gene and retrieve pieces of information about the regulatory motif
Pieces of information to retrieve
Mitores Entry Accession number
Mitores Entry product description
Mitores Gene name
UTRef Accession number
UTRef UTR Type
UTRSite accession number
UTRSite Standard name
Involved Databases
MitoRes
UTRef
UTRSite

10. TUTORIAL: a simple case study involving federated DB (2/4)
UTRef
MitoRes
UTRSite
GenBank
OMIM
PUBMED
HmtDB
Uniprot
EMBL_CDS

11. TUTORIAL: a simple case study involving federated DB (3/4)
UTRef
MitoRes
UTRSite

12. TUTORIAL: a simple case study involving federated DB (4/4)
Need
Studying the regulation of the expression of the topoisomerase I at mRNA level.
Biologist needs to retrieve information about regulation of expression of the topoisomerase I. To this end she decides to investigate if there are UTR sequences responsible for the regulation of the expression of the corresponding gene and retrieve pieces of information about the regulatory motif
Pieces of information to retrieve
Under the cover
Mitores Entry Accession number
Mitores Entry product description
UTRef Accession number
UTRef Type
UTRSite accession number
UTRSite Standard name
Involved Databases
MitoRes
UTRef
UTRSite
SQL Query
SELECT DISTINCT
"t1"."DESCRIPTION" AS "Description", "t2"."MITONUC_ID" AS "ID", "t3"."UTRDB_ID" AS "UTRef ID", "t3"."TYPE" AS "UTR type", "t4"."NAME" AS "Gene name", "t5"."UTRSITEID" AS "UTRSite ID", "t5"."STANDARDNAME" AS "Standard name"
FROM "LIBI"."MITONUC_GENE" "t2" LEFT JOIN "LIBI"."MITONUC_GENE_PRODUCT" "t6" ON "t2"."GENE_ID" = "t6"."GENE_ID"
RIGHT JOIN "DDQB"."MITONUC_PRODUCT_DDQB" "t1" ON "t6"."PRODUCT_ID" = "t1"."PRODUCT_ID" LEFT JOIN "LIBI"."MITONUC_GENE_MRNA" "t7" ON "t2"."GENE_ID" = "t7"."GENE_ID" RIGHT JOIN "DDQB"."MITONUC_MRNA_DDQB" "t8" ON "t7"."MRNA_ID" = "t8"."MRNA_ID" LEFT JOIN "DDQB"."MITONUC_UTR_VIEW" "t3" ON "t8"."MRNA_ID" = "t3"."MRNA_ID" LEFT JOIN "LIBI"."UTREF_UTR" "t9" ON "t3"."UTRDB_ID" = "t9"."ACCESSION" LEFT JOIN "LIBI"."UTREF_SIGNAL" "t10" ON "t9"."ACCESSION" = "t10"."ACCESSION" LEFT JOIN "DDQB"."UTRSITE_UTRSITE_DDQB" "t5" ON "t10"."UTRSITEID" = "t5"."UTRSITEID" LEFT JOIN "LIBI"."MITONUC_GENE_GENE_NAME" "t11" ON "t2"."GENE_ID" = "t11"."GENE_ID" RIGHT JOIN "LIBI"."MITONUC_GENE_NAME" "t4" ON "t11"."GENE_NAME_ID" = "t4"."GENE_NAME_ID" WHERE UPPER("t1"."DESCRIPTION") LIKE '%TOPOISOMERASE I%'

13. Modeling a data abstraction in bioinformatics: DDQB
IBM Data Discovery and Query Builder (DDQB) is a powerful search tool with a graphical interface that enables users with various levels of expertise to easily configure queries and leverage the full spectrum of information assets.
By means of DDQB researchers can query the federated DB not in term of its physical fields, but in term of more abstract entities arranged into taxonomies that have been specifically developed for LIBI users, so queries become tasks closer to their research subjects than to informatics activities.
For the DDQB tutorial: