P336: Automating CMC Dossier Creation from End-to-End with automated Data to Content
Poster Presenter
Jayashree Tikhe
Senior Product Marketing Manager
Dassault Systemes United States
Objectives
The objective of this study was to examine how we would use a combination of knowledge graphs, a structured authoring environment and generative AI to transform CMC authoring to be automated and data driven.
Method
We took a sample set of stability data from a laboratory system & generated a knowledge graph which was used to automate queries from a structured authoring environment & generate the resulting tables and figures. Using generative AI summarization & conclusions we generated a full stability section.
Results
We were successful in pipelining the data and were able to generate the appropriate charts and the human readable view of the content. Generative AI was good at producing the appropriate content and narrative. We restricted the scope to preparing the content and narrative, so there was a human review to completion. However, there are opportunities to fully automate this in the future.
Conclusion
We need to take full advantage of the methodologies and capabilities around knowledge graphs, structured authoring environments and generative AI to transform CMC solution to deliver accurate and efficient CMC documentation. This is a key component to transform the industry approach to CMC generation; it is no longer sufficient to create documents in word. Combining structured and unstructured data in this way enables use of transmission standards like FHIR from the same data.
