Towards Efficient Use of Image Resources in Oryzabase

Oryzabase is an integrated rice science database that is characterized by strain resources and a publication-based gene dictionary. In spite of having more than 7000 featured image resources stored in the database, it is difficult to look at an overview of all images because each image is usually buried under a deep hierarchy of web pages. In order to make use of all the data efficiently, we started developing an image retrieval system, called Gallery.

Although all images in the database belong to a resource (wild collection, induced mutants, embryo mutants/lethal embryo mutants, enhancer trap lines, genes, developmental stage), neither keyword nor explanation was directly linked to each image data. Therefore first we assigned all images to some of the eight anatomical terms such as (1) leaf, (2) grain, (3) flower, (4) chlorophyll, (5) embryo, (6) inflorescence/spikelet, (7) plant body, (8) lesion mimic, regardless of the resource it belongs to. Starting with one of the above eight categories, users can further narrow down their search by the resource, and finally reach the strains, genes and/or developmental stages information they wanted. If we could apply plant ontology to these images, potentiality of image data would be greatly increased. 

We implemented Apache Lucene/Solr as a search platform for the Gallery.Oryzabase also uses the same solr technology to provide a refine search for references. Oryzabase is available at http://www.shigen.nig.ac.jp/rice/oryzabaseV4/.