Active learning-driven optimization of gelation time for cell morphology control in synthetic hydrogels

Hydrogels are widely used for cell encapsulation due to their soft, porous structure, tunability, and high-water content. Dynamic network properties, which more closely mimic the time-dependent variation of tissues in vivo, are often achieved by adding synthetic polymers and crosslinking them under external stimuli. While one-pot synthesis simplifies this process, it requires a deeper understanding of how gelation time influences cell morphology. Since reaction kinetics are highly sensitive to temperature and pH, phase diagrams correlating reaction conditions with gelation time can guide cell assembly. We adopt an active learning approach to optimize response surface predictions using Gaussian Process Regression (GPR), and improve gelation time prediction at physiological pH with prior knowledge at neutral pH, where literature is available. Our findings show that gelation time reliably predicts the aspect ratio of encapsulated cells. Furthermore, we confirm that cell shape is influenced by the properties of the forming networks as cells establish connections with the matrix early on, by inhibiting focal adhesion kinase. This work highlights the potential of high-throughput microrheology in facilitating the fabrication of synthetic extracellular matrices.