Mougin Fleur
ERIAS, INSERM U897, ISPED, University of Bordeaux, Bordeaux, France.
Stud Health Technol Inform. 2015;210:195-9.
Significant efforts have been undertaken for providing the Gene Ontology (GO) in a computable format as well as for enriching it with logical definitions. Automated approaches can thus be applied to GO for assisting its maintenance and for checking its internal coherence. However, inconsistencies may still remain within GO. In this frame, the objective of this work was to audit GO relationships. First, reasoning over relationships was exploited for detecting redundant relations existing between GO concepts. Missing necessary and sufficient conditions were then identified based on the compositional structure of the preferred names of GO concepts. More than one thousand redundant relations and 500 missing necessary and sufficient conditions were found. The proposed approach was thus successful for detecting inconsistencies within GO relations. The application of lexical approaches as well as the exploitation of synonyms and textual definitions could be useful for identifying additional necessary and sufficient conditions. Multiple necessary and sufficient conditions for a given GO concept may be indicative of inconsistencies.
人们已经做出了巨大努力，以可计算的格式提供基因本体论（GO），并通过逻辑定义对其进行扩充。因此，自动化方法可应用于GO，以协助其维护并检查其内部一致性。然而，GO中可能仍然存在不一致之处。在此框架下，这项工作的目标是审核GO关系。首先，利用关系推理来检测GO概念之间存在的冗余关系。然后根据GO概念首选名称的组成结构，识别缺失的必要和充分条件。发现了一千多个冗余关系和500个缺失的必要和充分条件。因此，所提出的方法成功地检测到了GO关系中的不一致之处。词汇方法的应用以及同义词和文本定义的利用，可能有助于识别更多的必要和充分条件。给定GO概念的多个必要和充分条件可能表明存在不一致。
