Oral Presentation Royal Australian Chemical Institute National Congress 2026

Synthesis of biomedical polyurethanes in flow guided by Bayesian reaction optimisation (136640)

Max J. H. Worthington 1 , Lewis Blackman 1 , Mark Bown 1 , Yanan Fan 2 , James Gardiner 1 , Veronica Glattauer 1 , Alex Shmaylov 1 , Tanja Junkers 3 , Timothy Hughes 1
  1. Biomedical Manufacturing, CSIRO, Clayton, VIC, Australia
  2. Data61, CSIRO, Eveleigh, NSW, Australia
  3. School of Chemistry, Monash University, Clayton, VIC, Australia

Medical devices generally require biologically inactive material for use in living creatures. Polyurethanes have proven a versatile class of polymers which can meet specific device requirements: non-toxic, tunable elasticity and hardness, and either biodegradability or stability in the body for decades1. Since the 1960s, many polyurethanes that fall under the prototypical reaction scheme have been developed, patented and locked into exclusive supply contracts. To allow continued material innovation, the discovery of novel polyurethanes is required. To meet this need, we have developed a platform to rapidly test and prototype new polyurethanes via flow synthesis. In a single experiment, reaction conditions are modified until a material with predefined chemical and physical properties is obtained via a Bayesian approach to experimental design. In this approach, an initial batch of polymer is synthesised, characterised, and the results processed by the optimiser to determine the next best step to reach the desired material. This process is continued stepwise until the desired material is obtained or the experimental budget is met. In using Bayesian optimisation, we can rapidly determine if a given set of reagents can afford a viable polyurethane for use in biomedicine, saving time over alternative classical Design of Experiments approaches2. The benefit of building this capability in a computer-controlled flow system allows us to work towards an autonomous and self-optimising material discovery system in the future3. This presentation will give an overview of our work, as well as broader efforts in this area from within CSIRO4, and give a brief tutorial on the implementation of Bayesian-optimisation guided experimental design using free and open-source software.

  1. Gunatillake, P. A.; Dandeniyage, L. S.; Adhikari, R.; Bown, M.; Shanks, R.; Adhikari, B. Advancements in the Development of Biostable Polyurethanes. Polym. Rev. 2019, 59 (3), 391–417. https://doi.org/10.1080/15583724.2018.1493694.
  2. Shields, B. J.; Stevens, J.; Li, J.; Parasram, M.; Damani, F.; Alvarado, J. I. M.; Janey, J. M.; Adams, R. P.; Doyle, A. G. Bayesian Reaction Optimization as a Tool for Chemical Synthesis. Nature 2021, 590 (7844), 89–96. https://doi.org/10.1038/s41586-021-03213-y.
  3. Rubens, M.; Vrijsen, J. H.; Laun, J.; Junkers, T. Precise Polymer Synthesis by Autonomous Self-Optimizing Flow Reactors. Angew. Chem. Int. Ed. 2019, 58 (10), 3183–3187. https://doi.org/10.1002/anie.201810384.
  4. Kohl, T. M.; Zuo, Y.; Muir, B. W.; Hornung, C. H.; Polyzos, A.; Zhu, Y.; Wang, X.; Alexander, D. L. J. Machine-Learning Assisted Optimisation during Heterogeneous Photocatalytic Degradation Utilising a Static Mixer under Continuous Flow. React. Chem. Eng. 2024. https://doi.org/10.1039/D3RE00570D.