School of Pharmacy, Faculty of Medicine, The Chinese University of Hong Kong, Shatin, New Territories, Hong Kong SAR.
School of Pharmacy, Faculty of Medicine, The Chinese University of Hong Kong, Shatin, New Territories, Hong Kong SAR
Drug Metab Dispos. 2022 Jan;50(1):86-94. doi: 10.1124/dmd.121.000420. Epub 2021 Oct 25.
An HERB-Drug Interaction (HDI) database is a structured data collection method for HDI information extracted from scattered literatures for quick retrieval. Our review summarized the ten currently available HDI databases, including those databases comprising HDI on the market. A detailed comparison on the scope of monographs, including the nature of content extracted from the original literature and user interfaces of these databases, was performed, and the number of references of fifty popular herbs in each HDI database was counted and presented in a heatmap to give users an intuitive understanding of the focuses of different HDI databases. Since it is well known that the development and maintenance of databases need continuous investment of capital and manpower, the sustainability of these databases was also reviewed and compared. Recently, artificial intelligence (AI) technologies, especially Natural Language Processing (NLP), have been applied to screen specific topics from massive articles and automatically identify the names of drugs and herbs in the literature. However, its application on the labor-intensive extraction and evaluation of HDI-related experimental conditions and results from literature remains limited due to the scarcity of these HDI data and the lack of well-established annotated datasets for these specific NLP recognition tasks. In view of the difficulties faced by current HDI databases and potential expansion of AI application in HDI database development, we propose a standardized format for data reporting and use of Concept Unique Identifier (CUI) for medical terms in the literature to accelerate the structured data collection. SIGNIFICANCE STATEMENT: The worldwide popularity of botanical and/or traditional medicine products has raised safety concerns due to potential HDI. However, the publicly available HDI databases are mostly outdated or incomplete. Through our review of the currently available HDI databases, a clear understanding of the key issues could be obtained and possible solutions to overcome the labour-intensive extraction as well as professional evaluation of information in HDI database development are proposed.
草药-药物相互作用（HDI）数据库是一种结构化的数据收集方法，用于从分散的文献中提取 HDI 信息，以便快速检索。我们的综述总结了目前可用的十个 HDI 数据库，包括包含市场上 HDI 的数据库。对这些数据库的专着范围进行了详细比较，包括从原始文献中提取的内容的性质和这些数据库的用户界面，并对每个 HDI 数据库中五十种流行草药的参考文献数量进行了计数，并以热图形式呈现，使用户直观地了解不同 HDI 数据库的重点。由于众所周知，数据库的开发和维护需要持续的资金和人力投入，因此还对这些数据库的可持续性进行了审查和比较。最近，人工智能（AI）技术，特别是自然语言处理（NLP），已被用于从大量文章中筛选特定主题，并自动识别文献中药物和草药的名称。然而，由于这些 HDI 数据的稀缺性以及缺乏针对这些特定 NLP 识别任务的完善标注数据集，其在从文献中提取和评估与 HDI 相关的实验条件和结果方面的应用仍然有限。鉴于当前 HDI 数据库面临的困难以及 AI 在 HDI 数据库开发中的潜在应用扩展，我们提出了一种用于数据报告的标准化格式，并在文献中使用概念唯一标识符（CUI）来加速结构化数据收集。意义陈述：由于潜在的 HDI，植物药和/或传统医学产品在全球范围内的普及引起了安全性问题。然而，现有的公开 HDI 数据库大多已过时或不完整。通过对目前可用的 HDI 数据库进行综述，可以清楚地了解关键问题，并提出可能的解决方案，以克服在 HDI 数据库开发中信息的劳动密集型提取和专业评估。
