Enhanced natural language access to anatomically-indexed data

Enhanced Free Text Access to Anatomically-Indexed Data Gail Sinclair Informatics Division University of Edinburgh 80 South Bridge Edinburgh, UK EH1 1HN carols@dai.ed.ac.uk Bonnie Webber Informatics Division University of Edinburgh 2 Buccleuch Place Edinburgh, UK EH8 9LW bonnie.webber@ed.ac.uk Duncan Davidson MRC Human Genetics Unit Western General Hospital Crewe Road Edinburgh, UK EH4 2XU duncan.davidson@hgu.mrc.ac.uk Abstract We describe our use of an existing re- source, the Mouse Anatomical Nomen- clature, to improve a symbolic interface to anatomically-indexed gene expression data. The goal is to reduce user effort in specifying anatomical structures of inter- est and increase precision and recall. 1 Introduction Language Technology (LT) resources are time- consuming and expensive to develop, and appli- cations rarely have the luxury of calling upon re- sources specially designed for the task at hand. For LT applications in developmental biology such as robust interfaces to anatomically-indexed gene ex- pression data and text mining tools to assist in build- ing such databases, resources already exist in the form of anatomical nomenclatures for several model organisms including mouse, zebrafish, drosophila and human. (Others may follow.) These nomen- clatures have been developed by biologists for biol- ogists, to record in a clear, intuitive and structured way the structures that can be distinguished at each stage of an embryo’s development. The challenge for LT applications in developmental biology is to stretch them to serve other purposes as well. In this paper, we describe how we have taken one of these anatomical nomenclatures (mouse) and ex- tracted from it a new resource to facilitate free text access to anatomically-indexed data. The techniques we have used are applicable to anatomical nomen- clatures for other model organisms as well. The paper is organised as follows: In Section 2, we describe the Mouse Atlas, which is the particular context for the interface we are developing. Sec- tion 3 describes what we are doing to reduce the amount of effort a user has to expend in specifying anatomical structures of interest to them. In Sec- tion 4, we describe how what we did to reduce user effort also serves to provide a clearer display of the results of searching. Then in Sections 5 and 6, we describe what we are doing to increase the precision and recall of user queries. 2 The Mouse Atlas The Mouse Atlas, developed by researchers at the Medical Research Council’s Human Genetics Unit (MRC HGU) in Edinburgh, is a 3D atlas of mouse embryo development (http://genex.hgu.mrc.ac.uk). Anatomical structures within each of the 26 Theiler Stages of embryo development are labelled, and 3D reconstructions of each stage can be displayed in transverse, frontal, sagittal or arbitrary planes. The Mouse Atlas is now being used to support in- dexing of gene expression data, allowing the results of gene expression experiments to be indexed with respect to where genes are expressed in the devel- oping embryo. There are at least two ways of using anatomy to index gene expression data. In spatial indexing, data is associated directly with volume el- ements, voxels …

Content truncated to 500 words