A method of searching for related literature on protein structure analysis by considering a user's intention

• Published in: BMC bioinformatics Vol. 16 Suppl 7; no. S7; p. S4
• Main Authors: Ito, Azusa, Ohkawa, Takenao
• Format: Journal Article
• Language: English
• Published: England BioMed Central 23.04.2015
• Subjects: Databases, Protein; Evaluation Studies as Topic; Humans; Information Storage and Retrieval - methods; Intention; Periodicals as Topic; Protein Conformation; Proteins - chemistry; User-Computer Interface
• ISSN: 1471-2105; 1471-2105
• DOI: 10.1186/1471-2105-16-s7-s4

In recent years, with advances in techniques for protein structure analysis, the knowledge about protein structure and function has been published in a vast number of articles. A method to search for specific publications from such a large pool of articles is needed. In this paper, we propose a method to search for related articles on protein structure analysis by using an article itself as a query. Each article is represented as a set of concepts in the proposed method. Then, by using similarities among concepts formulated from databases such as Gene Ontology, similarities between articles are evaluated. In this framework, the desired search results vary depending on the user's search intention because a variety of information is included in a single article. Therefore, the proposed method provides not only one input article (primary article) but also additional articles related to it as an input query to determine the search intention of the user, based on the relationship between two query articles. In other words, based on the concepts contained in the input article and additional articles, we actualize a relevant literature search that considers user intention by varying the degree of attention given to each concept and modifying the concept hierarchy graph. We performed an experiment to retrieve relevant papers from articles on protein structure analysis registered in the Protein Data Bank by using three query datasets. The experimental results yielded search results with better accuracy than when user intention was not considered, confirming the effectiveness of the proposed method.