Developing the strains and processes at the Lab.

Problem description

Design-Build-Test-Learn (DBTL) cycles have been widely applied to solve engineering tasks. “Self-driving laboratories” are a next level of DBTL that allow to optimize certain properties of a system of interest in an iterative and fully automated manner.

Once an initial design has been made in-silico, it can be implemented and tested on an automated robotics platform. The thereby generated data are fed to an AI/ML framework that predicts the best design of experiments for the next iteration. A “Selflearning laboratory system” requires well-defined and well-managed workflows, seamless data and information flows between the individual phases, programmatic access to equipment and advanced data analytics across scales and types.

References

Goal

To develop autonomous systems that:

  • evaluate and develop ‘’self-learning” experimental design algorithms;
  • combine data-driven approaches with mechanistic modeling;
  • optimize experimental DBTL algorithms;
  • apply the developed methods to DBTL use cases in biotech (i) method, (ii) strain and/or (iii) process development.

Project team

Joery de Vries
Joery de Vries
TU Delft
Matthijs Spaan
Matthijs Spaan
TU Delft
Willi Gottstein
Willi Gottstein
DSM

Other projects

Our partners & members