DSM partners on AI biomanufacturing as part of €2.5M move

12 Jan 2021 --- DSM is partnering with Dutch university TU Delft to establish the Artificial Intelligence Lab for Biosciences (the AI4B.io Lab). According to the company, this lab will be the first of its kind in Europe to apply AI to full-scale biomanufacturing, from microbial strain development to process optimization and scheduling.

DSM plans to invest €2.5 million (US$3.0 million) into the laboratory over the first five years. The lab is currently hiring PhD students and anticipates that its first results will be seen in early 2022.

“The lab is envisioned to speed up innovation, for example through the development and application of highly-accurate AI-based models of microbes, bioprocesses and the combination of both,” Hans Roubos, corporate science fellow of biological sciences at DSM, tells NutritionInsight.

Fermentation at its core
Roubos explains that biomanufacturing uses biological systems to produce commercially important biomaterials, biomolecules and ingredients. These can be used in food and beverage processing, medicines and industrial applications.

“The focus of the AI4B.io Lab will be on fermentation processes, including upstream (raw materials), fermentation units as such, and downstream processing (purification of product for desired quality).”

The lab will also explore AI methods from microbe-level to a full factory scale, with applications ranging from strain development and application development to fermentation process control and scheduling.

Novel AI applications will, for example, be demonstrated in the biomanufacturing of enzymes and dairy cultures. “Thanks to the increased innovation speed, more leads for interesting products can be explored, and production processes can be optimized early on,” says Roubos. Integrating biosciences and digital technologies can help reduce the time spent on innovation cycles, from prototyping to scaling and commercialization.

Digital twins guide the way
Traditionally, scientific research is based on trial and error within multiple sub-studies that work together toward a specific objective, such as a new product or production technology. However, AI allows scientists to invert this process. 

The technology creates “digital twins” or a virtual mirror of the desired real-world situation. Machine learning can then be used to help determine how to achieve this. 

While AI is already widely applied in engineering research – for instance, to replace physical wind turbines or tunnels with digital twins – DSM observes gaps in biosciences and biotechnology.

“AI is already commonly used for self-driving cars and the control of robots. Differences with full-scale processes in biomanufacturing are the limited amounts and quality of available on-line sensor data that represent the behavior of the microbes and processes versus those in cars and robots,” explains Roubos.

Furthermore, real-life experimentation with cars and robots can be performed cost-efficiently, while such experiments in biomanufacturing are infeasible due to the cost factor. Additionally, they may produce a lot of waste, which is at odds with the environment. 

Optimizing R&D
AI can help biomanufacturing by finding patterns in data to guide optimization processes in R&D. This could be in automated strain development and screening projects, for example.

“Another aspect is the speed-up of calculations, where neural networks can be trained to resemble calculations of mechanistic models, but are computationally ten to 100 times more efficient,” says Roubos. 

Such a speed-up allows for real-time optimization and adaptive steering of processes based on the digital twins, taking into account microbial behavior in a full-scale bioreactor while simulating flow patterns and oxygen and sugar gradients and acting on process-sensor data.DSM hopes to maximize its impact by linking fundamental research to concrete opportunities. 

Bio-economy capital of the world
The AI4B.io Lab will be part of the Dutch National Innovation Center for AI (ICAI), which works to keep the Netherlands at the forefront of knowledge and talent development in AI. 

The partnership is also touted as strengthening Delft’s status as the bio-economy capital of the world.

The AI4B.io Lab will be the third ICAI Lab on the TU Delft campus, joining the AI for Retail Lab Delft of Ahold Delhaize, and the AI for Fintech Lab of ING. 

TU Delft will also invest in 24 interdisciplinary AI laboratories on a broad range of topics to further drive collaboration between scientists working in AI and scientists from other domains.

Additionally, The AI4B.io Lab will also collaborate with Planet B.io, the open-innovation ecosystem at the Biotech Campus Delft. This could be by providing research insights and consultancy to biotechnology start-ups on the campus, for example. Both DSM and TU Delft are founding partners of Planet B.io.

A string of partnerships
This marks the latest in a string of DSM partnerships. Last month, the company closed a transaction with biotechnology company Amyris. Under the terms of the agreement, Amyris is licensing DSM rights to assume the supply of farnesene – a bio-based compound – to Givaudan for the production and sale of a single specialty ingredient.

In November, DSM teamed up with Huami Corp to enhance its personalized nutrition arm and develop technologies for health monitoring. 

Additionally, DSM recently announced a partnership with a professional road cycling team previously known as Team Sunweb. 

By Katherine Durrell