Argudas - argumentation-based data sharing across gene expression databases

Modern biological and medical science is producing large amounts of experimental data. Much of this data is held in public databases and accessible via the Internet to researchers across the world. The content of these databases does partially overlap. Due to the variation of experimental details and conditions, and sometimes simply due to human error during the handling of this data, it is now quite common to find some inconsistencies between these databases. For a researcher making use of these databases it poses the question of which of the data sources to trust and how to determine who is right and who is wrong. As part of this project we propose to use so-called argumentation technology to deal with this problem. Just as humans present arguments in favour and against a particular statement and how a discussion can move forward and backward between two people, computers can carry out similar reasoning if they are provided with the underlying knowledge to do so. In this project, we therefore propose to develop a tool that can use data from such public databases and domain expert knowledge to argue over the inconsistencies within and between these biomedical databases. The output of our tool will show the biologists what statements are supported by which databases and how they relate to each other. This will allow the biologists to come to an informed decision with respect to which data he or she can have most confidence in. The proposed work is focusing on a specific set of such databases which hold gene-expression data, i.e. data that captures which genes are active in which parts of an animal or human.

Due to experimental variations, differences in result interpretation and occasional human error during the execution of experiments and the handling of the data they produce, it is now a common fact that there are inconsistencies within and across biomedical databases, and that there is currently no systematic, computational support for end users to deal with these inconsistencies. The primary objective of this project is the development and deployment of a web-based argumentation tool that allows end users to explore differences in a number of key gene expression databases, covering in-situ and microarray gene expression and, possibly, SAGE data. Although the work is focusing on gene expression, general lessons will be learnt, preparing the possible introduction of argumentation-based data sharing in other biomedical application domains. Argumentation, a form of computer-based reasoning where arguments in favour and against a particular statement are considered, has been shown to work in medical situations, where it can be used to provide both decision support to medics and explanations of those decisions to patients. Although supported by a clear theoretical basis, argumentation follows a natural form of reasoning that is easily accessible and, hence, more readily acceptable by scientists without training in formal logic and mathematics. Our previous work has shown that this form of reasoning is also suitable to address inconsistencies and incompleteness within and across biomedical data resources. While our previous work has established much of the technical groundwork, the purpose of this project is to take argumentation technology from the laboratory of an informatics group to the biologist end user. To this end we particularly emphasise the evaluation with end users and the focus on two use cases at the MRC Human Genetics Unit and the European Bioinformatics Institute. The completed Argudas system will be hosted at the MRC after this project completes.