Artificial Intelligence (AI) is a hot topic - so in industrial biotechnology, but how can AI be implemented and does it drive innovation? These questions were selected by members of the industrial biotechnology association IWBio e.V. during a workshop on April 1st, 2025 hosted by board member Martin Langer (BRAIN Biotech AG). Together with board members the event was co-organized by the IWBio members Dr. Anja Schwenzfeier (AB Enzymes) and Peter Kitschmann (Bausch & Ströbel). Perspectives on the implementation and impact of AI along the innovation value chain - from contract research and manufacturing organizations to legal advice in this field - were discussed.
Klaus Mauch set the scene during a keynote vividly explaining computing power and data sizes handled by today’s AI tools and defining AI as a provider of correlations where the human mechanistic world ends. He shared advantages and disadvantages of using cloud-based systems or in-house solutions for data management as well as potential costs for implementation of AI in daily business. Implementation of AI can be challenging due to complexity of the technology, lacking expertise of employees, and sometimes the missing explainability of results. The critical amount and quality of data used to train AI was a challenge echoed by the workshop participants.
Mauch impressively demonstrated that AI is not just a trend but is already present in our daily lives. Its use will speed up and optimize biotechnological R&D as well as production processes leading to a market advantage for companies considering the implementation as a strategic topic.
key insights from the workshop speakers
AI tools that bypass patents and set new benchmarks
Dr. Ingmar Schuster (Exazyme GmbH) shared his experience with tools for scientific literature searches & proprietary developments for novel protein engineering. Current benchmarks outperforming human performance have been shown for using free (e.g. PaperQA2) and paid versions (e.g. Elicit) of AI tools for scientific review writing - making literature search much easier. He showcased own developed tools and combination of tools that help design novel enzymes beyond directed evolution and rational design methods. Minimizing the risk of infringement of existing patents (less than 80% sequence similarity) was achieved by own developed inverse folding tools. Open-source options are RFDiffusion and Protein MPNN.
Virtual fermentation platform: design, process and maintenance
Dr. Martin Kessler (Wacker Chemie AG) elaborated on three potential use cases for AI support currently under investigation with partners.
- Saving time for fermenter design: AI is trained with data to reduce the number of necessary computational fluid dynamics (CFD) simulations, taking high precision but time consuming CFD simulations into account.
- Increase yield and efficiency by AI supported process control: real process data are used to train AI and finally steer standardized production processes, like timepoint of harvest.
- In maintenance, highly complex equipment is currently analyzed using sensors to prevent failures in future, e.g. by efficient maintenance.
AI in manufacturing – finding the best recipe
How data can be provided by real experiments and CFD simulation to train AI for filling optimization was shown by Maximilian Burger (Bausch & Ströbel). Unwanted air bubble inclusion during the injection of a liquid in a vial was used as a readout to generate CFD data and real data about the quality of the filling process. Together with machine learning using ensemble neuronal networks the optimal parameter for the liquid filling process – the recipe – shall be predicted in future.
IP regulation of AI products
Dr. Florian Reiling elaborated on the patentability of AI systems and tools in general, as well as the protectability of AI-generated results and processes. While AI systems and tools may be eligible for (software) patent protection, such protection is likely only available in very specific cases. Conversely, AI-generated results and processes are rather difficult to protect, as both copyright and patent protection necessitate human contribution, which is typically absent in the case of AI-generated works (however, trade secret protection may be a feasible option).
Participants of the workshop were made aware of the importance of carefully selecting training data and its sources. Risks may arise from external and/or internal data which (i) is subject to third-party rights and/or (ii) faces other chain-of-title issues. Dr. Reiling suggested that adapted license management and employee training could serve as measures to minimise these risks.
Dr. Reiling also emphasised that European AI regulation was adopted in August and 2024 and has entered in force in February 2025. Implementation will be carried out through a staggered process until 2030. Monitoring compliance with these obligations is fundamental in reducing related risks.
Where implementation of AI in biotechnology still falls short
While AI’s potential is undeniable, the participants also shed light on critical challenges that must be addressed before it becomes a seamless part of biotechnological workflows:
- Data quality & availability: AI is only as good as the data it learns from. Poor or incomplete datasets can lead to unreliable predictions, making it essential to integrate high-quality experimental and process data.
- Regulatory & IP risks: Landscape of intellectual property (IP) and patents in is evolving – risks of legal uncertainties around AI-generated discoveries and data mining have to .
- Implementation complexity: AI is not a plug-and-play solution, success depends on strategically integrating AI, investing time and money for future market advantage.
CONCLUSION:
During the IWBio workshop, it was confirmed that the use of AI in combination with machine learning is already a reality along the industrial biotechnology value chain. It also became clear that data quality, regulatory clarity and professional integration into existing processes are essential for efficient implementation. The workshop participants agreed that companies that use AI strategically will have a market advantage in the future - also in terms of their innovations.
Text and images: BRAIN Biotech AG
