M.A. Webb
Princeton University,
United States
Keywords: Bayesian optimization, active learning, rheology, macromolecule
Summary:
Designing polymers with targeted structural and functional properties remains a grand challenge in materials science. Nature achieves remarkable complexity and function through biopolymers' composition and interactions. Replicating such complexity in synthetic systems—or repurposing biopolymers for new applications—requires navigating an immense combinatorial design space, defined by unit chemistries, compositional heterogeneity, and architectural variations. This complexity limits rational design and complicates screening due to characterization challenges and data scarcity. However, the rise of machine learning, particularly in the context of the fifth paradigm of scientific discovery, offers promising new approaches. In this talk, I will discuss our recent efforts to integrate simulation, machine learning, and theory to map and navigate structure–function relationships in chemically and topologically diverse polymeric materials. I will emphasize strategies for overcoming data limitations through physics-informed models and algorithmic innovations. Finally, I will highlight key challenges and opportunities for machine learning in macromolecular characterization and simulation, illustrated through vignettes spanning biomolecular condensates, polymer solutions, and protein mimetics. These case studies will showcase both methodological advances and exciting applications at the intersection of polymer science, simulation, and artificial intelligence.