Falling Walls: Generative AI And the Development of Advanced Therapies

In an expert panel at the Falling Walls Science Summit, co-hosted by Sartorius, thought leaders from industry and research discussed the applicability of generative artificial intelligence in the development of modern medicine.

_{^{This article is posted on Sartorius Blog.}}

Moderator Mariette DiChristina from Boston University and Sartorius' Head of Advanced Data Analytics Johan Trygg were joined by Allison Duettmann, President of the Foresight Institute, David Ruau, Head of Strategic Alliances, Drug Discovery AI EMEA at NVIDEA, Mads Nørregaard-Madsen, Scientific Director of Amgen Copenhagen and Monika Lessl, Executive Director of the Bayer Foundation and Senior VP of Bayer. 

The panelists addressed the potential, challenges, and risks associated with the use of AI, particularly in the field of biologics, discussing the current state of generative AI, the scarcity of optimized health and technical data, and the complex nature of applying AI to multidisciplinary processes.

Generative AI has the potential to transform how medicines are made today. This is a huge opportunity, because we are able to analyze and understand biological and molecular data at scale.

Johan Trygg, Head of Advanced Data Analytics at Sartorius, full professor at Umeå University and visiting professor at Imperial College London.

AI could help design complex structures in molecule design

The panelists acknowledged the complexity of the topic and the fact that established AI models for use with natural language, such as ChatGPT, won't be transferable seamlessly for the use in therapy development. However, they were positive about the first steps being taken with generative AI, especially in areas like molecule design and for streamlining administrative tasks for example in clinical trials.

Some people argue that we should soon have a ChatGPT type of tool for drug development, where one can write in a prompt to come up with a molecule with very specific properties. We have to remain realistic - we are far off from that.

Johan Trygg, Head of Advanced Data Analytics at Sartorius, full professor at Umeå University and visiting professor at Imperial College London.

Availability and quality of data are crucial factors

According to the experts, a major hurdle is that existing data does often not have the quality and characteristics needed for AI use and is sometimes not available to everyone. On the one hand, this concerns patient data, which must be handled sensitively and requires good governance. On the other hand, a greater availability of technical data and documentation could help accelerate the implementation of AI in therapy development.

Wanted: Experts at the interface of data and biology

A key takeaway from the discussion was the need for highly skilled experts with an interdisciplinary understanding of data science and biology, who can bridge the gap between these disciplines. While currently rare, individuals with this kind of interdisciplinary training in multiple fields are expected to play an increasingly important role in the future of therapy development.