To Home page


The increasing amount of biological data being made available on-line, combined with the already vast numbers of research papers available electronically, makes Bioinformatics an exciting area for the application of Artificial Intelligence techniques.


Now that many genome sequencing efforts are complete, the emphasis is moving to interpreting the data. Gene expression networks, which describe genes and their interaction, are one such means of explanation, and on-going research in AIAI is applying Genetic Algorithms to search the space of potential gene expression networks, and identify the most promising. Early results are encouraging as experts judge the networks as plausible.


Ontologies are already being used and developed by biologists. The Gene Ontology is the best known example. Many anatomies of model species such as mouse, drosophila and C Elegans are also being published. These ontologies and anatomies are being used to index gene expression data. Building on these resources the XSPAN project is defining cross-species mappings which represent the links between homologous tissues. Integrating these resources will be a valuable contribution to the e-scientist exploring the developmental links between species. The XSPAN project has produced the COBrA Ontology Browser, which is an ontology browser and editor for GO and OBO ontologies. COBrA has been specifically designed to be usable by biologists to create links between ontologies, and has the following features:

  • allows drag-and-drop editing of GO ontologies
  • supports translation to OWL and other Semantic Web languages
  • supports the manual creation of mapings between terms in two ontologies.
COBrA was developed by Roman Korf. The modelling of is_a and part_of in anatomy ontologies is also being studied, and results are presented in this PSB paper More technical details of the OWL and RDF Schema can be found here.

Text Analysis

The retrieval and analysis of scientific texts is an important service. Current keyword-based approaches are limited, and new techniques are needed to generate mark-up in a machine interpretable form (in RDF, for example). In recent research, Inductive Logic Programming has been applied to learn information extraction rules which locate instances of ontology relations in texts. This supervised learning approach requires only a small set of annotated texts to generalise from. Further details can be found in this paper.

Projects Publications Contact
Hybrid GA
COBrA Ontology Browser
Aitken, Webber and Bard PSB 04
Learning IE Rules
Inferring Gene Networks from Microarray Data
Stuart Aitken
John Levine

to top
Updated: Fri Mar 19 18:24:07 GMT 2004
To the Edinburgh University website