OAMiner: Integrative Knowledge Anchored Hypothesis Discovery

描述(申请人提供)：该提案旨在解决生物学和医学领域对集成知识发现日益增长的需求，其目的是通过利用多个概念性知识来源，实现对不同数据集中基于图像的、表型的和生物分子特征之间相互关系的假设的发现、验证和验证--最终支持“高通量”的知识驱动的翻译科学。为了提供一个可管理的项目范围，骨关节炎倡议(OAI)数据集将被用作开发和评估计划研究产品的主要激励用例。该项目必然涉及主题专家(SME)对初始假设的分析，以进行系统培训和核查。然而，我们提出的方法的最终目标是将识别或验证知识锚定的假设的人工干预需求降至最低。为了产生这样的假设，使用了四个相互关联的知识来源：1)通过传统文本挖掘和对Medline数据库和相关全文资源库中找到的文章的NLP分析来获取全文发表的生物医学文献；2)国家医学图书馆统一医学语言系统(UMLS)中包含的公共可用本体；3)一个或多个数据库，包含表型和功能(例如，生活质量、心理、力量和表现指标)数据；以及4)计算机化图像分析得出的特征(例如，股四头肌的横截面面积)。

DESCRIPTION (provided by applicant):The objective of this proposal, which is designed to address the ever increasing need for integrated knowledge discovery in biology and medicine, is to enable the discovery, verification, and validation of hypotheses concerning interrelationships between image-based, phenotypic, and bio-molecular features in heterogeneous data sets by leveraging multiple conceptual knowledge sources - ultimately supporting "high throughput" knowledge-driven translational science. To provide for a manageable project scope, Osteoarthritis Initiative (OAI) data sets will be used as a primary, motivating use case for the development and evaluation of the projected research products. This project necessarily involves analysis of initial hypotheses by subject matter experts (SMEs) for system training and verification. However, the ultimate goal of our proposed approach is to minimize the need for human intervention to identify or validate knowledge-anchored hypotheses. In order to generate such hypotheses, four interrelated knowledge sources are used: 1) full-text published bio-medical literature accessed by both conventional text mining and NLP analyses of articles as found in the Medline database and associated full text repositories; 2) publically available ontologies included in the National Library of Medicine's Unified Medical Language System (UMLS); 3) one or more databases containing phenotypic and functional (e.g. quality of life, psychological, strength and performance measures) data; and 4) computerized-image analysis derived features (e.g. cross-sectional area of the quadriceps).