|
Bioinformatics is such a field where data grows at an exponential rate and knowledge grows only at linear rate. The ultimate challenge for the biological database community is to help in closing this gap between growth of data and knowledge. Bioinformatics data consists of different views of biological information. Bioinformatics databases are diverse in terms of data formats, and are highly redundant too. The data views in bioinformatics includes biological sequences (like DNA, RNA and proteins), gene or protein expression, functional properties, molecular interactions, clinical data, system descriptions, and related publications. The data appear as sequences, sequence annotations, structural models, physical maps, clinical records, interactions pathways, gene and protein expressions, protein-protein interactions, and other forms in data sources such as databases, private data collections, and related publications. Data warehousing emerges in bioinformatics to support biological knowledge discovery. A data warehousing is a structured repository of a large volume of data integrated from different operational sources to support analytical processing.
There is substantial diversity and variation in bioinformatics data, even among databases containing data of the same view (same type of data). Each database has its own infrastructure and proprietary data format i.e. common data standards and data exchange formats are not established in this field. For example, sequence entries are described in different formats in GenBank, Swiss-Prot, and EMBL. ASN.1 format (Abstract Syntax Notation One) is developed by GenBank while Swiss-Prot designed its own format. Also, the Swiss-Prot format slightly differs from the format of EMBL database. Further, the introduction of XML (Extensible Markup Language) as the generic data exchange format has also given rise to several variants of XML representations of bioinformatics data.
The bioinformatics data is characterized by enormous diversity matched by high redundancy, across individual as well as multiple databases. Enabling interoperability of the data from different sources requires resolution of data disparity and transformation in the common form (data integration), and the removal of redundant data, errors, and discrepancies (data cleaning). Frequently encountered data redundancy issues are:
- fragments and potential entries of the same item ,example sequence, may be stored in several source records
- databases update and cross-reference one another with a negative side effect of occasionally creating duplicates, redundant entries, and proliferating errors
- the same sequence may be submitted to more than one database without cross-referencing those records
- the owners of the sequence record may submit a sequence more than once to the same database
To enable the extraction of knowledge in a data warehousing environment, these are rectified by data warehousing integration and data cleaning components.
A bioinformatics data warehouse requires several components for operation:1) retrieval of data from databases, 2) mechanism for cleaning data, 3) flexibility of manipulating the datasets, and 4) integrating and designing purposeful analysis tools that can be used jointly or independently.
Currently, not all data warehousing concepts have been applied to bioinformatics. For example, the dimensional data model, based on relational tables, is not widely seen in bioinformatics due to the complexity of the real data. Data warehousing has been historically developed using mainly relational databases systems, which are not as broadly used in bioinformatics. The major components of a data warehouse provide for initial and incremental data integration, data annotation, and data mining. Data integration also includes subcomponents for data retrieval, data cleaning, and data transformation. These components enable compilation of raw data from various databases. Data cleaning support tools are used for filtering of irrelevant and redundant records. To enable the interoperability of tools on the heterogeneous datasets and for ease of data management, the data is transformed into a common data format or interoperable formats. Because of information being updated in databases on regular basis, these steps of retrieving, cleaning, and transforming must be repeated for an incremental data integration process.
Once the initial dataset is created, the annotation component enables the addition of value-added information from experimental results or related publications to the dataset. Data cleaning is also supported at the annotation stage, facilitating removal of erroneous data propagated from the data sources. The analysis of data is enabled by incorporating general or specific tools for data mining. Then the new knowledge extracted from these analyses is produced as an elaborate report.
|
