Skip to main content

Table 1 A comparison of five common data storage technologies currently deployed in annotation systems.

From: Next generation models for storage and representation of microbial biological annotation

  Free Text Tab/Line Delimited XML RDF/XML Relational-DB
Description Logic (FOL) NO NO NO YES NO
Ontology Standards NO NO NO YES NO
Centralized/not scalable NO YES* NO NO YES**
Human Readable YES YES NO YES YES
Domain Expert Understandable YES YES NO YES NO
Data Structure NONE Single Table Tree Graph Relational Tables
Data Expectations NONE NONE Schema - Constraints Inference rules Schema – Constraints
Native Format Text Text Text Text Binary
Query Engine Language Programmed by hand Programmed by hand Libraries available SPARQL SQL
Naming Standard NO UNA**** NO UNA NO UNA NO UNA UNA
Sequence Storage Solution In Text In Text XML/Indexed XML/Indexed Indexed
Search Speed (Worst Case) NP-Hard O(n) O(n) P O(log n)
Update Speed (Worst Case) NP-Hard O(n) O(n) P O(log n)
Conversion to Semantic Data loss possible – done by hand No data loss – done by hand No data loss – done with robust libraries - No data loss –library usage and some added labeling by hand
Conversion from Semantic No data loss Data loss Data loss - Data loss
  1. Free text is used in repositories such as scientific journals. Tab/Line delimited files are used in popular formats such as FASTA, GFF, and BLAST. Tab/Line delimited files also constitutes the bulk of program output from most bioinformatics software. Mature tools and sequence repositories such as GenBank support XML output. Many OWL based ontology repositories exist for semantic data integration, however RDF/XML data is currently scarce. Relational databases typically do not provide direct access to the data, instead a programming interface is provided for access to the underlying database. Free text is the most flexible, and also the least machine-readable. Relational databases are the most formal structures with the fastest indexing and searching capabilities. Relational technology requires the greatest computational expertise investment while free text is the most natural. XML and RDF/XML are designed for modification over time and in sharing data. In the rows discussing search speed and update speed, O(log n), O(n), P and NP-Hard are computer science terms indicating a range of how fast a computer solution can be obtained to a particular problem. P indicates a reasonable solution is possible in polynomial time, NP-Hard means that the solution space explodes relative to the input size. NP-Hard problems are expected to not be solvable on a computer in reasonable time. O(log n), O(n), and P are all solvable efficiently on a computer. In the rows discussing conversion to and from RDF/XML, Turtle, and other semantic aware data storage technologies, loss of information includes schema, constraints, data and formatting. For example, to convert from a relational schema to tab-delimited files, information is lost because the schema, triggers and views are not representable using tab-delimited files. So these columns are more than just data, they are data and descriptions surrounding the data for making logical conclusions and for executing computer codes in reasonable time. In the conversion from free text to semantic standards, assumptions and domain expertise may be lost.
  2. *Assuming all information is in one file. If multiple files exist, then an indexing system needs to be developed to organize information.
  3. **Relational databases are assumed to exist as a single installation on a powerful single resource. New database technologies have lessened this restriction in recent years.
  4. ***CWA – Closed World Assumption, missing information treated as false. OWA – Open World Assumption, missing information treated as unknown.
  5. ****UNA Unique Name Assumption – Each individual has a single unique name.