Designing Ca/Sr-Containing Glass Composition for Neutral pH-Responsive Dissolution Using AI

Objectives: Aiming at the development of smart bioactive restoratives, we are designing bioactive glasses that exhibit different functions by releasing various ions in response to surrounding pH. Previously, we developed calcium (Ca) and/or strontium (Sr)-containing glasses designed to exhibit low solubility in acidic environments and high solubility in neutral conditions to promote tooth remineralization. However, whether these compositions fully optimize pH-responsive dissolution behavior remains unclear. This study aimed to optimize Ca/Sr-containing glass composition to enhance its neutral pH-responsive dissolution using artificial intelligence (AI).
Methods: Fourteen glasses containing phosphorus (P), Ca, Sr, and gallium (Ga) were fabricated using the melt-quenching method. Their solubility was assessed after 24 hours of immersion in buffer solutions at six pH levels (4.50, 5.50, 6.37, 6.75, 7.12, and 7.50) for 24 hours. A Gaussian process regression model was trained on regularity in 14 glass compositions and 51 solubilities at six pH levels. Bayesian optimization explored 19405 possible combinations with mean probability scores. The candidate glass identified by the model was experimentally fabricated, and its solubility was evaluated. To quantify pH-responsive dissolution behavior, the solubility ratio (pH 7.12 / pH 5.50) was calculated and statistically compared between the candidate glass and the initial 14 glasses.
Results: Four candidate compositions were identified via Bayesian optimization. Among four candidates, the composition with the highest mean probability score (0.170) was selected as the optimized glass. The solubility of the optimized glass was 4.26% and 35.61% at pH 5.50 and 7.12, respectively. The solubility ratio (pH 7.12 / pH 5.50) of the optimized glass demonstrated significantly greater than those of the initial 14 glasses (p<0.05, ANOVA, Tukey's HSD test, n=3).
Conclusions: The optimized Ca/Sr-containing glass exhibited distinguished neutral pH-responsive dissolution behavior. These results highlight AI as a powerful tool for refining glass compositions to achieve pH-responsive dissolution, supporting its potential for developing smart bioactive materials.