(515g) Digital Assets, Digital Threads, and Digital Twins for Rapid Commercialization of Medicines: A Reference Architecture and Implementation to Enable Pharma 4.0

Today’s biopharma digital infrastructure is obsolete and fragmented. ELNs are expensive vaults with poor data engineering. PLMs are non-existent or they lack the breadth and depth expected of a true lifecycle management application. While Enterprise Data Lakes have had some success, they can be fragile and are not fully democratized to reach citizen data scientists. Project Portfolio Management may barely exist beyond glorified MS Excel/Project files. Data Science’s maximum expression is Random Forests. Machine Learning (e.g., image analysis and NLP) has not reached mainstream use, and In Silico Modeling results may not exist beyond an obscure and fragile shared folder. High Throughput Experimentation’s data is untapped due to the drag imposed by data wrangling. Ontologies are missing, rendering end-to-end analyses a mammoth undertaking. Reports, dashboards, may be more commonplace, but GxP validation can be a major burden. With this current landscape, the Pharma 4.0 revolution seems a very daunting task even for the most daring digital innovators in the industry. Finally, cultural norms and organizational constraints can make the transformation even more challenging.

In this work we present a future-view rooted in Amgen’s Process Development vast experience in this digital transformation space, and we share with the industrial community, regulatory agencies, and software vendors the many lessons learned. We envision a future of FAIR data assets that are curated and stewarded via effective data governance practices and processes. We imagine a digital thread that fetches these data assets from their sources, and changes them transactionally and programmatically with full traceability and scalability. We recognize that digital twins born out of machine learning and first principles modeling techniques will create cycle times near their theoretical minimum thanks to ubiquitous computation and augmented intelligence. We hypothesize a near-future where these connected assets, threads, and twins will revolutionize CMC authoring, quality processes, and enable major manufacturing and supply efficiencies. Finally, we identify a set of skills, methods, and practices to deliver this vision.