Machine-learning-driven rational design of gelatin architectures: saccharide derivatives as structural plasticizers under low water content.

Saccharide derivatives are highly promising, sustainable additives for tuning the physical profiles of gelatin-based green biomaterials. Although they are utilized in several applications in medical and food sciences, their structural roles under highly dehydrated conditions remain poorly understood. Here, we investigated the influence of various saccharide derivatives on gelatin networks across a broad structural spectrum, from bulk hydrogels to low-moisture sheets. Rheological measurements of the hydrogels revealed that the addition of saccharide derivatives elevated the gelation temperature and facilitated the network formation by restricting free water through preferential hydration. Intriguingly, this behavior inverted in the dried sheet state, where ATR-FTIR spectroscopy revealed that the saccharides disrupted the gelatin network to function as molecular plasticizers, significantly enhancing flexibility and extensibility. While residual water retained by saccharide hydration mainly drives this plasticization as previous study predicted, water amount alone fails to account for notable mechanical variations; for example, sorbitol-doped sheets required significantly less residual water to achieve equivalent elasticity compared to sucrose systems. To address this complexity, we leveraged a machine-learning-based Random Forest Regression (RFR) model. The RFR successfully decoupled the overlapping effects of water content and molecular features, identifying molecular volume and hydrogen-bond donor of the additives as the governing factors of mechanical properties. This predictive, data-driven framework offers new insights for the rational design of eco-friendly gelatin biomaterials with precisely tailored mechanics.