期刊
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
卷 12, 期 6, 页码 679-701出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TEVC.2008.915995
关键词
Conceptual clustering; database annotation; evolutionary algorithms (EAs); gene expression profiles; gene ontology (GO); knowledge discovery; multiobjective (MO) optimization
资金
- Spanish Ministry of Science and Technology [TIC-2003-00877, BIO2004-0270E, TIN2006-12879]
- Howard Hughes Medical Institute
- Washington University School of Medicine, St. Louis, MI
Current tools and techniques devoted to examine the content of large databases are often hampered by their inability to support searches based on criteria that are meaningful to their users. These shortcomings are particularly evident in data banks storing representations of structural data such as biological networks. Conceptual clustering techniques have demonstrated to be appropriate for uncovering relationships between features that characterize objects in structural data. However, typical conceptual clustering approaches normally recover the most obvious relations, but fail to discover the less frequent but more informative underlying data associations. The combination of evolutionary algorithms with multiobjective and multimodal optimization techniques constitutes a suitable tool for solving this problem. We propose a novel conceptual clustering methodology termed evolutionary multiobjective conceptual clustering (EMO-CC), relying on the NSCA-II multiobjective (MO) genetic algorithm. We apply this methodology to identify conceptual models in structural databases generated from gene ontologies. These models can explain and predict phenotypes in the immunoinflammatory response problem, similar to those provided by gene expression or other genetic markers. The analysis of these results reveals that our approach uncovers cohesive clusters, even those comprising a small number of observations explained by several features, which allows describing objects and their interactions from different perspectives and at different levels of detail.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据