Automated Integration of Biomedical Knowledge

今天，本体论是生物医学研究人员的重要工具，特别是在那些需要掌握大量信息和系统方法来设计和解释实验的领域，如癌症研究。事实上，本体论在生物医学研究的所有领域都在激增，既带来了挑战，也带来了机遇。这一领域的主要挑战之一源于这样一个事实，即本体是孤立地发展起来的，因此不可能从基因转移到生物体、疾病到药物。国家生物医学本体论中心（NCBO）代表了生物医学本体论的收集、协调和分布方面的基本努力，并提供了一个无与伦比的机会，将这些生物医学本体论结合到一个单一的搜索空间中，在这个搜索空间中，遗传、解剖、分子和药理学信息可以无缝地探索和利用，作为生物医学知识的整体表示。不幸的是，使用标准的人工管理方法进行本体集成是一项劳动密集型任务，无法扩大规模并跟上当前生物医学本体的增长速度。我们开发了一个基于信息论的自动化本体工程的系统框架，并成功地将其应用于基因本体（GO）的分析和工程，开发基因和蛋白质数据库，以及疾病进展和药物反应的外周生物标志物的识别。该项目汇集了一组独特的能力，包括本体工程、统计信号处理、生物信息学、癌症研究和临床药物基因组学，以信息理论的数学为基础，开发一种有原则的方法，自动组合和集成生物医学本体，并将其作为NCBO架构的一部分实施

Today, ontologies are critical instruments for biomedical investigators, especially in those areas, such as cancer research, that require the command of a vast amount of information and a systemic approach to the design and interpretation of experiments. In fact, ontologies are proliferating in all areas of biomedical research, offering both challenges and opportunities. One of the principal challenges of this field stems from the fact that ontologies are developed in isolation, rendering it impossible to move, for instance, from genes to organisms, to diseases, to drugs. The National Center for Biomedical Ontology (NCBO) represents a fundamental endeavor in the collection, coordination and distribution of biomedical ontologies and offers an unparalleled opportunity to combine these biomedical ontologies into a single search space where genetic, anatomic, molecular and pharmacological information can be seamlessly explored and exploited as a holistic representation of biomedical knowledge. Unfortunately, ontology integration using standard means of manual curation is a labor intensive task, unable to scale up and keep up with the current growth rate of biomedical ontologies. We have developed a systematic framework for automated ontology engineering based on information theory, and we have successfully applied it to the analysis and engineering of Gene Ontology (GO), the development gene and protein databases, and the identification of peripheral biomarkers of disease progression and drug response. This project brings together a unique group of competences, ranging from ontology engineering, statistical signal processing, bioinformatics, cancer research, and clinical pharmacogenomics, to develop a principled method, grounded on the mathematics of information theory, to automatically combine and integrate biomedical ontologies and implement it as part of the NCBO architecture