Digital Twins

Overview

Software models that mirror real bioprocessing operations in real time, enabling predictive process control, experimentation reduction, and learning from both process and product-level data. Michelangelo Canzaneri defines a digital twin analogously to Google Maps — a human-machine interface that mimics reality, predicts outcomes (estimated time of arrival), and guides decisions to avoid known failure modes. Merck's digital twin for CMC manufacturing required data layers, ontologies, models, and deep bioprocess expertise, but once built reduced experimentation by 50% per molecule and was designed for group-wide standardization rather than single-use-case deployment. (Segment 4 (15:00-20:00))