Description: Among regulatory professionals who support biopharmaceutical product development, it is common to use regulatory precedent (i.e., historical product approvals) to help inform strategic planning efforts. While there are many commercial service-providers who offer databases of historical approval information, these commercial databases typically contain a limited number of data fields that cannot be expanded based on each customer’s unique needs. Because of this, many regulatory professionals build and maintain their own lists and databases of historic approvals to address the specific needs of their organization and products. In this session, we will walk through a case study in building an extensible database of regulatory precedent by training machine learning software to read and extract information from FDA approval letters, approved drug labels, and FDA approval announcements.
Learning Objectives:
Learn about real-world applications of machine learning technology for regulatory affairs professionals
Understand the benefits and limitations of using software to extract data from unstructured documents like FDA approval letters, labels and announcements
Discuss human-in-the-loop systems and why regulatory professionals won’t be replaced by machine learning technology for a long, long time