This study was undertaken to systematically assess the utilities and performance of ontology-based semantic analysis in adverse outcome pathway (AOP) research. With an increasing number of AOPs developed by scientific domain experts to organize toxicity information and facilitate chemical risk assessment, there is a pressing need for objective approaches to evaluate the biological coherence and quality of these AOPs. Powered by ontologies covering a wide range of biological domains, abundant phenotypic data annotated ontologically, and some sophisticated knowledge computing tools, semantic analysis has great potential in this area of application. With the events in the AOP-Wiki first annotated into logical definitions and then grouped into phenotypic profiles by individual AOPs, the coherence and quality of AOPs were assessed at several levels: paired key event relationships (KER), all possible event pair combinations within AOPs, and the phenotypic profiles of AOPs, genes, biological pathways, human diseases, and selected chemicals. The semantic similarities were assessed at all these levels based on a unified cross-species vertebrate phenotype ontology encompassing the logical definitions of AOP events as well as many other domain ontologies. A substantial number of KERs and AOPs in the AOP-Wiki were found to be semantically coherent. These same coherent AOPs also mapped to many more genes, pathways, and diseases biologically aligned with the intended chain of events therein leading to their respective adverse outcomes. Significantly, these findings imply that semantic analysis should also have utilities in developing future AOPs by selecting candidate events from either the existing AOP-Wiki events or a broader collection of ontology terms semantically similar to the molecular initiating events or adverse outcomes of interest. In addition, semantic analysis enabled AOP networks to be constructed at the level of phenotypic profiles based on similarities, complementing those based on event sharing by bringing genes, pathways, diseases, and chemicals into the networks too-thus greatly expanding the biological scope and our understanding of AOPs.

中文翻译：

---

不良结果途径的语义特征。


本研究旨在系统地评估基于本体的语义分析在不良结果途径（AOP）研究中的效用和性能。随着科学领域专家开发出越来越多的 AOP 来组织毒性信息并促进化学风险评估，迫切需要客观的方法来评估这些 AOP 的生物一致性和质量。语义分析在涵盖广泛生物领域的本体、本体注释的丰富表型数据以及一些复杂的知识计算工具的支持下，在该领域具有巨大的应用潜力。 AOP-Wiki 中的事件首先注释为逻辑定义，然后按各个 AOP 分组为表型概况，AOP 的连贯性和质量在多个级别进行评估：配对关键事件关系 (KER)、AOP 内所有可能的事件对组合，以及 AOP、基因、生物途径、人类疾病和选定化学物质的表型特征。基于统一的跨物种脊椎动物表型本体（包含 AOP 事件的逻辑定义以及许多其他领域本体）对所有这些级别的语义相似性进行了评估。 AOP-Wiki 中的大量 KER 和 AOP 被发现在语义上是一致的。这些相同的连贯 AOP 还映射到更多的基因、途径和疾病，这些基因、途径和疾病在生物学上与其中导致各自不良结果的预期事件链一致。 值得注意的是，这些发现意味着语义分析也应该在开发未来的 AOP 中发挥作用，方法是从现有的 AOP-Wiki 事件或更广泛的本体术语集合中选择候选事件，这些本体术语在语义上类似于分子起始事件或感兴趣的不良结果。此外，语义分析使AOP网络能够在基于相似性的表型谱水平上构建，通过将基因、途径、疾病和化学物质带入网络来补充基于事件共享的网络，从而极大地扩展了生物学范围和我们的研究范围。对 AOP 的理解。