Use of synthetic data in AI development
Insights from a case study in Cryo-EM image analysis
IT
Abstract
Artificial Intelligence (AI), and in particular Machine Learning (ML) methods, unlock new approaches to digitalization through advanced data analysis – for instance for image processing. At the same time, novel high-resolution methods for collecting image data are opening up new use cases in analytical and quality control applications, like the use of cryogenic electron microscopy (cryo-EM) in pharma and biotechnology.
In the context of these developments, a collaboration project has been initiated between the company ATEM Structural Discovery and the GAMP D-A-CH Working Group on the GxP compliant use of AI. This project deals with the accurate and precise characterization of adeno-associated viruses (AAV) for quality control purposes via high-resolution cryo-EM nano-imaging and subsequent automated image analysis. AAV vectors are primarily used in advanced treatment options such as gene therapies or in personalized medicine applications – a growing area of innovation in pharma and biotechnology.
This article combines the use of cutting edge technology applied in analytical and quality control applications with requirements and challenges in GxP regulated environments considering process design, availability of data and generation of synthetic data as well as testing the robustness of the models. Thereby, this study is one practical example to apply Pharma 4.0 ideas that enable innovation via the interplay of advanced technology and a comprehensible, quality-oriented development process.
Abstract
Artificial Intelligence (AI), and in particular Machine Learning (ML) methods, unlock new approaches to digitalization through advanced data analysis – for instance for image processing. At the same time, novel high-resolution methods for collecting image data are opening up new use cases in analytical and quality control applications, like the use of cryogenic electron microscopy
